Rewrite kothic in Rust

[meenbeese]

Rust rewrite of kothic

This PR replaces the Python kothic MapCSS parser/processor and drules generator
with a complete Rust implementation. The behavior and output are byte-identical
to the Python reference; the native drawing-rules wire format is unchanged.

What changed

• The whole pipeline is now native Rust: kothic (single style), gen_all
  (all 6 styles), merge_variants and merge_styles (family/merged files).
• No Python interpreter, no runtime dependencies beyond libc — the binaries are
  self-contained (LTO release builds, ~3 MB each).
• All Python-isms from the port were removed: CamelCase methods renamed to
  idiomatic snake_case (updateStyles → update_styles, testChains →
  test_chains, parseZoom → parse_zoom, addZoom/addCondition/addStyles
  → add_*, …), whatever_to_rgb/hex/cairo → parse_color_*, Python-flavored
  helpers (Number() → number(), py_list_repr → list_repr,
  komap_mapswithme → generate_drules), dead code dropped (empty newGroup(),
  unused TOKEN_SUBPART, the unused Parser._src field), and the
  #![allow(non_snake_case)] banners are gone.

Speed

On my personal machine, generating all 6 Organic Maps styles (gen_all):

	time	peak RSS
Python	~14.6 s	~109 MB
Rust	~2.4 s	~47 MB

≈ 6× faster than Python, and peak memory usage is less than half of before.

Verified byte-identical output

Rust and Python were both run against the real Organic Maps data/ tree:

• All 6 single-variant *.bin and *.txt drules files: identical.
• Side files types.txt, classificator.txt, visibility.txt, colors.txt,
  patterns.txt: identical.
• The full generate_drules.sh flow (single-style → merge_variants →
  merge_styles) reproduces the committed
  drules_default/outdoors/vehicle.{bin,txt}, drules_design.bin and
  drules_merged.{bin,txt} byte-for-byte.

Tests / lint

• cargo build clean, cargo test: all 65 tests pass.
• cargo clippy --all-targets only produces 3 warnings.

[meenbeese]

PTAL @biodranik

[biodranik]

What is the motivation for the rewrite? Only the speed? Is speed now the main issue with Kothic?

Python code is easier to maintain/validate by humans. I like Rust, but we don't have enough experts at the moment to maintain/review these changes efficiently. How good is your experience with Rust and this Kothic code?

[meenbeese]

My motivation with the rewrite was to mostly improve the architecture, but the speed improvement doesn't hurt at all. I find this Rust version way easier to understand and review, and we can also split into more modules as needed. I think for a tool like this, it makes more sense to use a fast and compiled language that produces a single binary.

[meenbeese]

Rebased on top of, and integrated the changes from master.

[meenbeese]

@biodranik It would be great to merge this PR (I will make a subsequent PR on the main repo to integrate the new version), as I want to pursue the open style issues and the Rust version is so much more easier to understand and work with, which is also the case for LLMs and agentic coding tools as well.

[biodranik]

I have doubts, because we don't have other contributors to do quality code reviews for vibecoded rust. Blindly trusting LLMs is not a great idea.

Which model/efforts did you use to port btw?

[meenbeese]

Well, I don't think converting to Rust suddenly makes the code more complicated. And it can be a learning opportunity as well :) I tried to keep the old structure as much as possible and didn't use too much idiomatic Rust.

I used the Codex cli fyi.