Cache the match object in the compiled re pattern

Feature or enhancement

Proposal:

re allocates a new match object for every successful match. Matches of one pattern all have the same size, and in a typical loop the previous match dies just before the next one is created.

Keep one dead match object in the compiled pattern and reuse it for the next match instead of allocating. This saves 15-18 ns per successful match (e.g. p.match('42') goes 89 ns → 71 ns, and the string-to-Fraction constructor gains ~6%), with no behavior change. The cache slot is accessed with atomic operations; validated with the test suite, refleak runs, and a concurrent stress test on the free-threaded build.

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Linked PRs

• gh-153162

从已编译的 re 模式实现开始，检查成功匹配是如何分配的。使用现有测试套件确认行为保持不变，然后在 free-threaded 构建上运行 refleak 检查和并发压力测试。当已失效的匹配对象能够安全地复用，并且报告的性能提升得以保持时，即视为完成。