Unify lock allocation and GIL release thresholds

[ifduyue]

Summary

Align the lazy lock allocation threshold with the GIL release threshold in _xxhash.c.

Changes

• src/_xxhash.c: XXHASH_LOCK_MAYBE_INIT allocated the lock at >= XXHASH_GIL_MINSIZE while every GIL release check uses >, so an update of exactly 64KB allocated a lock that was never used. Both now use > XXHASH_GIL_MINSIZE
• README.rst: update the thread-safety wording to match ("more than 64KB", "operations of 64KB or less")
• Bump version to 4.0.0.dev8

Testing

• Boundary behavior verified at 64KB-1 / 64KB / 64KB+1 (results match the one-shot equivalents)
• Full test suite passes (132 tests)

[codspeed-hq]

Merging this PR will not alter performance

✅ 148 untouched benchmarks
⏩ 66 skipped benchmarks¹

---

_{Comparing fix/unify-gil-threshold (f0b2b5b) with master (b274f52)}

Footnotes

1. 66 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩