Introduce Array API

[RandomNameUser]

Added cleaner APIs for array handling (find first, best, all matches)
Added GIL release for array ops
Added docs for array ops to Readme
Added tests for array ops
Added explicit hamming_distance_byte to have the same interface as hamming_distance_string. Functionally the find first array op does the same thing, but this being Python I'm weary of having two apparently interchangeable functions that have different return types, that just smells like trouble. ;)

It all seems to work fine, as far as I can tell. The only thing I don't understand is the benchmarking: the best and all array ops are about 2 orders of magnitude faster than the first one. Either the compiler does very high-level optimization magic, or something is wrong that I can't see...

Thanks!

[mrecachinas]

This looks fine. I only noted two changes before we'll want to merge. I'm not certain off-hand why you're observing that benchmark behavior, but I'd have to look further into it.

I'm comfortable merging and cutting a release after you make those changes if it suits your needs. 👍

[Copilot]

Pull Request Overview

This PR adds cleaner APIs for array handling by introducing three new functions for finding matches within arrays of byte data. The changes include GIL release for array operations, comprehensive test coverage, documentation updates, and benchmarking support for the new functionality.

Key changes:

• Added three new array operation functions: first, best, and all matches within a specified Hamming distance
• Introduced check_bytes_within_dist function for byte-to-byte comparison with explicit naming
• Added GIL release mechanisms for improved performance in multi-threaded environments

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 6 comments.

File	Description
test/test_hexhamming.py	Comprehensive test coverage for new array APIs with edge cases and benchmarking
hexhamming/python_hexhamming.cc	Implementation of new wrapper functions with GIL release and improved error handling
hexhamming/_version.h	Version bump to 2.3.0 reflecting new functionality
README.rst	Documentation for array APIs with examples and usage tips

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[mrecachinas]

Hey, I dug into the benchmark anomaly -- turns out it was a combination of bugs that we've now fixed:

1. Py_BuildValue("i") was silently truncating the int64/uint64 return values to 32-bit C ints, which could produce garbage values that confused the benchmark harness
2. The best function had an overflow bug (best_dist = max_dist + 1) that, after finding the first match, set the threshold to -1 — causing it to inadvertently skip the early-exit check and compute raw distances for every remaining element

After fixing those (plus a malloc safety fix and a NEON edge case), benchmarks now show the expected ordering: first is fastest (early exit), then all, then best (has to scan everything twice to find the minimum). Numbers look right -- first is ~5-18x faster than the others depending on the test case.