Fix simplification of short antimeridian crossings

[shkyyy18]

Bug

simplify() projects longitude by subtracting the start longitude directly. For a short straight track with longitudes 179.9, 180, -179.9 at latitude 0, that treats the endpoint displacement as -359.8 degrees instead of +0.2 degrees. The intermediate collinear point is consequently retained even with a 5 metre tolerance. The same problem occurs for westbound crossings and either spelling of the dateline (180 or -180).

Fix

Normalize each local longitude displacement to [-pi, pi) before applying the existing equirectangular projection. This keeps the current small-track projection and public API, while handling the longitude seam correctly. This does not claim to support globally winding tracks or change the existing projection's large-extent limitations.

Regression coverage and validation

• Four synthetic straight-crossing cases (east/west, +/-180) fail on upstream main and pass with the fix.
• Two additional route cases ensure a significant bend across the dateline remains intact.
• Input point lists remain unchanged.
• Full suite: 902 passed, tested locally on Python 3.12 and 3.13.
• All configured prek run --all-files hooks passed on Python 3.13, including Ruff, the five type checkers and generated CLI documentation.
• Updated changelog and README per contributing guidance. No new dependencies or CI changes.

Prepared with AI assistance; the regression failure and full validation were run against a local checkout using synthetic coordinates only.

[sgraaf]

Hi! Thank you for this bug fix. I noticed there was a misplaced line in the README, which I have removed. Other than that: LGTM!