release: v0.2.4 - fix trailing_zeros regression

- Remove fast-size-class feature flag entirely
- Restore match-based size_to_class (12% regression fix)
- Update benchmarks: AethAlloc wins multithread churn at 17M ops/s

v0.2.3 - Benchmark suite and documentation

- Add comprehensive BENCHMARK.md with full methodology
- Add bulk_init optimization for magazine caching
- Fix clippy warning (unsafe bulk_init)
- Update benchmark results vs glibc, jemalloc, mimalloc, tcmalloc
- Multithread churn: 19.4M ops/s (#1)
- Tail latency P99: 106ns (tied #1)
- Fragmentation RSS: 17MB (#1, 1.8x better than glibc)

v0.2.2 - Documentation update

- Update BENCHMARK.md version references to 0.2.1
- Full benchmark suite vs glibc, jemalloc, mimalloc, tcmalloc
- Multithread churn: 19.4M ops/s (#1)
- Tail latency P99: 106ns (tied #1)
- Fragmentation RSS: 17MB (#1, 1.8x better than glibc)

v0.2.1 - Benchmark updates and documentation

- Add comprehensive BENCHMARK.md with full methodology
- Add bulk_init optimization for magazine caching
- Fix clippy warning (unsafe bulk_init)
- Update benchmark results vs glibc, jemalloc, mimalloc, tcmalloc
- Multithread churn: 19.4M ops/s (#1)
- Tail latency P99: 106ns (tied #1)
- Fragmentation RSS: 17MB (#1, 1.8x better than glibc)

Merge pull request #1 from shift/feature/stress-tests

feat: add stress tests and fix critical bugs