Result of turning off debuginfo in dev/test/bench, adding thinLTO to bench

a.txt

Compile-time measurements after turning off debuginfo for dev/test/bench,
and turning on thinLTO for bench in https://github.com/tikv/tikv/pull/4689

The dev/test profiles are affected by
https://github.com/tikv/tikv/issues/4711, where RocksDB's debuginfo is
not being turned off.

commit1=8d69c7a94e282d26c46a7cc5bf4fe9ac8e22cb15
commit2=8001ee0e5a5a38ed0eb3d10f2282f82f820a936e

dev full: 10.5%
dev partial: 17.5%
test full: 12.4%
test partial: 10.3%
bench full: -22.5%
bench partial: -37.59%

Raw data:

cargo build full before
    Finished dev [unoptimized + debuginfo] target(s) in 7m 37s
cargo build partial before
    Finished dev [unoptimized + debuginfo] target(s) in 40.59s
cargo build full after
    Finished dev [unoptimized] target(s) in 6m 49s
cargo build partial after
    Finished dev [unoptimized] target(s) in 33.46s
cargo test --no-run full before
    Finished dev [unoptimized + debuginfo] target(s) in 33m 53s
cargo test --no-run partial before
    Finished dev [unoptimized + debuginfo] target(s) in 2m 55s
cargo test --no-run full after
    Finished dev [unoptimized] target(s) in 29m 40s
cargo test --no-run partial after
    Finished dev [unoptimized] target(s) in 2m 37s
cargo bench --no-run full before
    Finished release [optimized + debuginfo] target(s) in 77m 27s
cargo bench --no-run partial before
    Finished release [optimized + debuginfo] target(s) in 39m 59s
cargo bench --no-run full after
    Finished release [optimized + debuginfo] target(s) in 94m 53s
cargo bench --no-run partial after
    Finished release [optimized + debuginfo] target(s) in 55m 01s