Add native SVG polyline path serializer and bump ABI to 36

[FarhanAliRaza]

Motivation

• Add a native, zero-copy SVG polyline path serializer to produce compact path strings in Rust instead of building one Python string per point for performance and consistency.
• Expose the serializer across the C ABI so Python can call into the Rust core and avoid per-point Python allocations.
• Bump the ABI version to reflect the new exported function so loaders can detect mismatched libraries.

Description

• Bump ABI version from 35 to 36 in both src/lib.rs and python/xy/_native.py to account for the new FFI entry.
• Implement svg::poly_path in a new src/svg.rs module that formats screen-space coordinates into an SVG path string with the same fixed-point formatting as the public SVG exporter.
• Add a C ABI entry xy_svg_poly_path in src/lib.rs that returns the required byte count and supports size-query retries, and wire its signature into the Python ctypes wrapper in python/xy/_native.py.
• Add a Python convenience wrapper svg_poly_path that calls the FFI with a resizable buffer, and switch _poly_path in python/xy/_svg.py to use _native.svg_poly_path.
• Add unit tests: a Rust test in src/svg.rs and a Python test test_native_svg_poly_path_format_and_validation in tests/test_kernels.py to validate formatting and error handling.

Testing

• Ran the Rust unit tests with cargo test, and the new svg::poly_path test passed.
• Ran Python unit tests with pytest, including the new tests/test_kernels.py::test_native_svg_poly_path_format_and_validation, and all tests passed.
• Verified the Python FFI path-resize/retry behavior by exercising the svg_poly_path wrapper in the added test.

---

Codex Task

[codspeed-hq]

Merging this PR will improve performance by ×2.7

⚡ 1 improved benchmark
✅ 93 untouched benchmarks
⏩ 1 skipped benchmark¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	test_svg_export_line	18.1 ms	6.6 ms	×2.7

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing codex/evaluate-svg-rasterization-speed-in-rust (f14122b) with main (e2d0460)}

Footnotes

1. 1 benchmark was skipped, so the baseline result was used instead. If it was deleted from the codebase, click here and archive it to remove it from the performance reports. ↩