store_subscr results with 3 specializations

STORE_SUBSCR3.md

3 Specialized opcodes:

* STORE_SUBSCR_LIST_INT
* STORE_SUBSCR_DICT_UNICODE
* STORE_SUBSCR_BYTEARRAY_INT

All benchmarks below are with GCC PGO on WSL, but without CPU isolation.

Pyperformance:

Benchmark	1025_main	store_subscr3
nbody	126 ms	106 ms: 1.19x faster
unpack_sequence	45.4 ns	41.9 ns: 1.08x faster
meteor_contest	110 ms	106 ms: 1.03x faster
logging_format	8.91 us	8.61 us: 1.03x faster
logging_simple	7.85 us	7.62 us: 1.03x faster
richards	65.7 ms	64.1 ms: 1.02x faster
logging_silent	123 ns	121 ns: 1.02x faster
json_loads	29.3 us	28.8 us: 1.02x faster
regex_compile	160 ms	158 ms: 1.02x faster
sympy_integrate	24.5 ms	24.2 ms: 1.01x faster
pickle_list	4.31 us	4.26 us: 1.01x faster
nqueens	97.2 ms	96.0 ms: 1.01x faster
scimark_lu	151 ms	150 ms: 1.01x faster
scimark_sor	167 ms	166 ms: 1.01x faster
hexiom	8.02 ms	7.96 ms: 1.01x faster
sympy_sum	198 ms	197 ms: 1.01x faster
python_startup_no_site	6.35 ms	6.36 ms: 1.00x slower
xml_etree_iterparse	111 ms	112 ms: 1.01x slower
unpickle_pure_python	298 us	301 us: 1.01x slower
regex_v8	23.4 ms	23.6 ms: 1.01x slower
xml_etree_process	62.5 ms	63.2 ms: 1.01x slower
dulwich_log	84.1 ms	85.2 ms: 1.01x slower
deltablue	5.21 ms	5.28 ms: 1.01x slower
pathlib	21.2 ms	21.5 ms: 1.02x slower
pickle_dict	27.8 us	28.2 us: 1.02x slower
pickle_pure_python	406 us	413 us: 1.02x slower
mako	12.5 ms	12.8 ms: 1.02x slower
chaos	87.8 ms	89.7 ms: 1.02x slower
scimark_sparse_mat_mult	5.96 ms	6.09 ms: 1.02x slower
pyflate	566 ms	579 ms: 1.02x slower
unpickle_list	5.12 us	5.25 us: 1.03x slower
json_dumps	13.4 ms	13.7 ms: 1.03x slower
scimark_fft	400 ms	414 ms: 1.03x slower
fannkuch	463 ms	479 ms: 1.03x slower
telco	6.73 ms	6.97 ms: 1.04x slower
chameleon	7.60 ms	7.99 ms: 1.05x slower
pickle	12.0 us	12.6 us: 1.05x slower
Geometric mean	(ref)	1.00x faster

Benchmark hidden because not significant (18): go, django_template, xml_etree_parse, raytrace, sympy_expand, regex_effbot, sympy_str, crypto_pyaes, python_startup, pidigits, regex_dna, 2to3, float, xml_etree_generate, tornado_http, scimark_monte_carlo, spectral_norm, unpickle

Microbenchmarks:

from pyperf import Runner
runner = Runner()

runner.timeit("dict[str]=...",
    setup="from itertools import repeat; d = dict()",
    stmt="for s in repeat('foobar', 100_000):\n"
         "    d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = s"
)
runner.timeit("dict[tuple]=...",
    setup="from itertools import repeat; d = dict()",
    stmt="for s in repeat((1, 2), 100_000):\n"
         "    d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = d[s] = s"
)
runner.timeit("list[int]=...",
    setup="from itertools import repeat; L = [0]*10",
    stmt="for i in repeat(2, 100_000):\n"
         "    L[i] = L[i] = L[i] = L[i] = L[i] = L[i] = L[i] = L[i] = L[i] = L[i] = i"
)
runner.timeit("bytearray[int]=...",
    setup="from itertools import repeat; a = bytearray([0]*10)",
    stmt="for i in repeat(2, 100_000):\n"
         "    a[i] = a[i] = a[i] = a[i] = a[i] = a[i] = a[i] = a[i] = a[i] = a[i] = i"
)

Benchmark	main_subscr_micro	subscr_micro
bytearray[int]=...	16.6 ms	8.14 ms: 2.03x faster
list[int]=...	15.6 ms	7.77 ms: 2.01x faster
dict[str]=...	18.5 ms	13.2 ms: 1.41x faster
dict[tuple]=...	23.5 ms	23.1 ms: 1.02x faster
Geometric mean	(ref)	1.56x faster

Specialization details:

Summary:
      bm_crypto_pyaes: 99.7%
   bm_django_template: 77.0%
          bm_fannkuch: 33.5%
    bm_meteor_contest: 99.9%
             bm_nbody: 99.98%
           bm_nqueens: 66.3%
     bm_regex_compile: 98.2%
         bm_regex_dna: 99.87%
          bm_regex_v8: 99.47%
           bm_scimark: 0.01% <-- array.array throws things off
  
        weighted mean: 46.9%
  mean of percentages: 77.4%


bm_crypto_pyaes: 99.7%
    store_subscr.specialization_success : 60
    store_subscr.specialization_failure : 47
    store_subscr.hit : 428697
    store_subscr.deferred : 616
    store_subscr.miss : 7
    store_subscr.deopt : 0
    store_subscr.unquickened : 363

bm_django_template: 77.0%
    store_subscr.specialization_success : 91
    store_subscr.specialization_failure : 1359
    store_subscr.hit : 290305 
    store_subscr.deferred : 83763
    store_subscr.miss : 119
    store_subscr.deopt : 0
    store_subscr.unquickened : 1200

bm_fannkuch: 33.5%
    store_subscr.specialization_success : 52
    store_subscr.specialization_failure : 108223
    store_subscr.hit : 3538352
    store_subscr.deferred : 6925551
    store_subscr.miss : 7
    store_subscr.deopt : 0
    store_subscr.unquickened : 353

bm_meteor_contest: 99.9%
    store_subscr.specialization_success : 51
    store_subscr.specialization_failure : 20
    store_subscr.hit : 1030974
    store_subscr.deferred : 571
    store_subscr.miss : 7
    store_subscr.deopt : 0
    store_subscr.unquickened : 337

bm_nbody: 99.98%
    store_subscr.specialization_success : 62
    store_subscr.specialization_failure : 20
    store_subscr.hit : 6001629
    store_subscr.deferred : 572
    store_subscr.miss : 7
    store_subscr.deopt : 0
    store_subscr.unquickened : 376

bm_nqueens: 66.3%
    store_subscr.specialization_success : 53
    store_subscr.specialization_failure : 6870
    store_subscr.hit : 878448
    store_subscr.deferred : 438965
    store_subscr.miss : 7
    store_subscr.deopt : 0
    store_subscr.unquickened : 349

bm_regex_compile: 98.2%
    store_subscr.specialization_success : 56
    store_subscr.specialization_failure : 540
    store_subscr.hit : 1866375
    store_subscr.deferred : 33863
    store_subscr.miss : 46
    store_subscr.deopt : 0
    store_subscr.unquickened : 329

bm_regex_dna: 99.87%
    store_subscr.specialization_success : 53
    store_subscr.specialization_failure : 20
    store_subscr.hit : 815135
    store_subscr.deferred : 596
    store_subscr.miss : 7
    store_subscr.deopt : 0
    store_subscr.unquickened : 337

bm_regex_v8: 99.47%
    store_subscr.specialization_success : 55
    store_subscr.specialization_failure : 34
    store_subscr.hit : 332275
    store_subscr.deferred : 1335
    store_subscr.miss : 13
    store_subscr.deopt : 0
    store_subscr.unquickened : 326

bm_scimark: 0.01%
    store_subscr.specialization_success : 50
    store_subscr.specialization_failure : 146953
    store_subscr.hit : 1666
    store_subscr.deferred : 9403966
    store_subscr.miss : 7
    store_subscr.deopt : 0
    store_subscr.unquickened : 407

Another run:

Benchmark	1026_main	1026_store_subscr
nbody	128 ms	102 ms: 1.25x faster
pickle_dict	30.4 us	28.4 us: 1.07x faster
hexiom	8.35 ms	8.03 ms: 1.04x faster
pickle_list	4.54 us	4.38 us: 1.03x faster
django_template	46.5 ms	45.5 ms: 1.02x faster
json_dumps	13.8 ms	13.5 ms: 1.02x faster
pidigits	180 ms	177 ms: 1.02x faster
fannkuch	468 ms	461 ms: 1.02x faster
sympy_integrate	24.5 ms	24.1 ms: 1.01x faster
meteor_contest	111 ms	109 ms: 1.01x faster
scimark_lu	152 ms	150 ms: 1.01x faster
unpack_sequence	40.5 ns	40.3 ns: 1.01x faster
regex_dna	200 ms	201 ms: 1.01x slower
python_startup_no_site	6.35 ms	6.40 ms: 1.01x slower
python_startup	9.02 ms	9.09 ms: 1.01x slower
mako	12.8 ms	12.9 ms: 1.01x slower
go	176 ms	178 ms: 1.01x slower
unpickle_pure_python	298 us	302 us: 1.01x slower
json_loads	28.7 us	29.2 us: 1.02x slower
xml_etree_iterparse	110 ms	112 ms: 1.02x slower
pickle_pure_python	409 us	417 us: 1.02x slower
dulwich_log	83.8 ms	85.6 ms: 1.02x slower
chameleon	7.65 ms	7.83 ms: 1.02x slower
regex_effbot	3.17 ms	3.24 ms: 1.02x slower
xml_etree_parse	153 ms	157 ms: 1.03x slower
deltablue	5.16 ms	5.33 ms: 1.03x slower
logging_silent	121 ns	126 ns: 1.04x slower
telco	6.74 ms	7.01 ms: 1.04x slower
pathlib	20.7 ms	21.7 ms: 1.05x slower
unpickle_list	4.95 us	5.26 us: 1.06x slower
Geometric mean	(ref)	1.00x faster

Benchmark hidden because not significant (25): pickle, tornado_http, crypto_pyaes, sympy_expand, nqueens, regex_compile, scimark_sparse_mat_mult, xml_etree_process, 2to3, sympy_sum, scimark_monte_carlo, logging_simple, raytrace, chaos, float, scimark_sor, xml_etree_generate, sympy_str, pyflate, spectral_norm, regex_v8, richards, scimark_fft, logging_format, unpickle