Speed up core logic using Rust

[jacobwhall]

This PR rewrites some of the core distance calculations in Rust, using crates such as kiddo (k-d trees) and rayon (parallelization) to faithfully re-implement this logic. The result is a drop-in replacement for the original, passing all tests and running significantly faster.

A few notes:

• I have used uv for dependency management and to run CI tests
• Maturin handles cross-platform compilation for package distribution
• Since I was already working in pyproject.toml, I went ahead and bumped the version to 0.4.0

I recommend releasing something like 0.4.0-alpha to ensure the package distributes correctly, and give end users the chance to test this package in their environments before making a full 0.4.0 release.

[jacobwhall]

I removed the progress_bar option, since it will need a reimplementation in Rust.

[jacobwhall]

Quick before/after benchmark:

Before:

Benchmark 1: uv run examples/simple_example.py
  Time (mean ± σ):      1.139 s ±  0.016 s    [User: 4.539 s, System: 0.093 s]
  Range (min … max):    1.116 s …  1.165 s    10 runs

After:

Benchmark 1: uv run examples/simple_example.py
  Time (mean ± σ):     199.9 ms ±   9.0 ms    [User: 1930.3 ms, System: 65.0 ms]
  Range (min … max):   186.7 ms … 215.1 ms    14 runs

Please take this with a grain of salt since benchmarks often struggle to represent real-world situations, and my computer might perform differently than yours.