[ruff] Migrate cache from bincode to rkyv

[epage]

Summary

bincode is unmaintained.
This isn't a problem on its own but if a problem comes up with it. As ruff's use of bincode is relatively isolated, the risk is low.

Options:

• defer: reevaluating what to do
• use a fork: didn't see one I was confident in
• fork: nothing pressing us to do this atm
• bitcode: faster than bincode but blocked on Make unsound code opt-in. SoftbearStudios/bitcode#51 but how that is being handled doesn't give me much confidence in the maintenance
• rkyv: almost as fast as bitcode and highly respected

If it wasn't for rkyv looking so promising, I would say we defer this.

Fixes #22284

Test Plan

Leveraging existing tests as this should be transparent

[charliermarsh]

We use rkyv heavily in uv!

[astral-sh-bot]

Typing conformance results

No changes detected ✅

Current numbers

The percentage of diagnostics emitted that were expected errors held steady at 94.47%. The percentage of expected errors that received a diagnostic held steady at 89.19%. The number of fully passing files held steady at 95/134.

[astral-sh-bot]

Memory usage report

Memory usage unchanged ✅

[charliermarsh]

Are you able to include some (brief) summarized numbers on cache size and performance? More concerned with the former, I trust the latter is similar.

[astral-sh-bot]

ecosystem-analyzer results

No diagnostic changes detected ✅

Full report with detailed diff (timing results)

[astral-sh-bot]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

Formatter (stable)

✅ ecosystem check detected no format changes.

Formatter (preview)

✅ ecosystem check detected no format changes.

[epage]

Are you able to include some (brief) summarized numbers on cache size and performance? More concerned with the former, I trust the latter is similar.

Do we have a typical set of use cases we run against?

[MichaReiser]

@ntBre do you remember which project's you used when you analyzed the caching/diagnostic rendering regression recently (I don't remember what it was).

I often used airflow and CPython

[ntBre]

I don't remember specifically, but I also usually use our cpython benchmark for this kind of thing.

[MichaReiser]

This is the issue I had in mind #18198 (comment), where you used airflow, homeassistant and Cpython

[epage]

Before

$ cargo build --release --bin ruff && hyperfine --warmup 10   "./target/release/ruff check ./crates/ruff_linter/resources/test/airflow/ --no-cache -e -s --isolated"
  "./target/release/ruff check ./crates/ruff_linter/resources/test/airflow -e -s --isolated"
warning: ignoring 'build.analysis' config, pass `-Zbuild-analysis` to enable it
   Compiling ruff v0.15.19 (/Users/epage/code/oss/ruff/crates/ruff)
    Finished `release` profile [optimized] target(s) in 1m 44s
Benchmark 1: ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow/ --no-cache -e -s --isolated
  Time (mean ± σ):     161.8 ms ±  10.7 ms    [User: 701.6 ms, System: 1186.6 ms]
  Range (min … max):   148.8 ms … 187.0 ms    18 runs

Benchmark 2: ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow -e -s --isolated
  Time (mean ± σ):      75.6 ms ±   1.9 ms    [User: 81.3 ms, System: 578.8 ms]
  Range (min … max):    73.6 ms …  80.8 ms    37 runs

Summary
  ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow -e -s --isolated ran
    2.14 ± 0.15 times faster than ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow/ --no-cache -e -s --isolated

$ rm -rf ruff_cache && cargo build --release --bin ruff && ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow/ -e -s --isolated && du -ch .ruff_cache
warning: ignoring 'build.analysis' config, pass `-Zbuild-analysis` to enable it
    Finished `release` profile [optimized] target(s) in 0.15s
2.6M    .ruff_cache/0.15.19
2.6M    .ruff_cache
2.6M    total

After

$ cargo build --release --bin ruff && hyperfine --warmup 10   "./target/release/ruff check ./crates/ruff_linter/resources/test/airflow/ --no-cache -e -s --isolated"
  "./target/release/ruff check ./crates/ruff_linter/resources/test/airflow -e -s --isolated"
warning: ignoring 'build.analysis' config, pass `-Zbuild-analysis` to enable it
    Finished `release` profile [optimized] target(s) in 0.18s
Benchmark 1: ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow/ --no-cache -e -s --isolated
  Time (mean ± σ):     160.5 ms ±  11.5 ms    [User: 700.9 ms, System: 1201.8 ms]
  Range (min … max):   146.4 ms … 180.1 ms    18 runs

Benchmark 2: ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow -e -s --isolated
  Time (mean ± σ):      71.6 ms ±   1.8 ms    [User: 81.2 ms, System: 579.2 ms]
  Range (min … max):    69.4 ms …  77.2 ms    36 runs

Summary
  ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow -e -s --isolated ran
    2.24 ± 0.17 times faster than ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow/ --no-cache -e -s --isolated

$ rm -rf .ruff_cache/ && cargo build --release --bin ruff && ./target/release/ruff check ./crates/ruff_linter/resources/test/airflow/ -e -s --isolated && du -ch .ruf
f_cache
warning: ignoring 'build.analysis' config, pass `-Zbuild-analysis` to enable it
   Compiling ruff v0.15.19 (/Users/epage/code/oss/ruff/crates/ruff)
    Finished `release` profile [optimized] target(s) in 1m 46s
2.7M    .ruff_cache/0.15.19
2.7M    .ruff_cache
2.7M    total