Prepare 4.0.0 release

[ifduyue]

Summary

Release preparation for 4.0.0, the next major version.

Changes

• Breaking: remove deprecated xxhash.VERSION_TUPLE (deprecated in v3.6.0, scheduled for removal in the next major release)
• Bump version to 4.0.0.dev0
• Document the thread-safety model and update the README version example
• Update the changelog for 4.0.0: drop Python 3.8, remove VERSION_TUPLE, upgrade xxHash to v0.8.3 (with a note about the upstream XXH64_update autovectorization regression on AVX2-targeted GCC/Clang builds; shipped wheels are unaffected), per-object locking for sub-interpreter and free-threaded Python support, pyodide wasm32 wheels, s390x test job, Python 3.15 classifier, and the sharded PyPI upload

Testing

• Full test suite passes (132 tests; free-threading/sub-interpreter cases skip on regular CPython)

[codspeed-hq]

Merging this PR will not alter performance

✅ 148 untouched benchmarks
⏩ 66 skipped benchmarks¹

---

_{Comparing prepare-4.0.0 (950ca16) with master (b751afa)}

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. ↩