Python: write stroke rate as cadence256 and rounded cadence.

[sanderroosendaal]

Developer field 93 and fractional_cadence are no longer written. 28.5 spm rounds half up to cadence 29.

Co-authored-by: Cursor cursoragent@cursor.com

Implements the Python side of #16. Should be merged after #16.

The diff against main contains only this change.

🤖 Generated with Claude Code

[JBlaak]

Two things before this goes in:

1. Precedence differs between writer and reader. _native_cadence prefers stroke_rate, resolve_stroke_rate prefers cadence256. With Record(stroke_rate=30, cadence256=28.5) the file gets 30 while validate() checks timing against 28.5. Could Record keep a single stroke-rate attribute (e.g. cadence256 in spm, written and read), or at least use the same order in both places?

2. The stroke-rate typical-range warning goes dead. validation.py only looks up _TYPICAL while looping over RECORD_DEVELOPER_ATTRS, and stroke_rate was mapped through field 93. With 93 removed, the "stroke_rate": (10.0, 100.0) entry is never reached. It needs checking against the resolved rate, plus a test.

[sanderroosendaal]

Both points addressed in 83f4431, which also merges main:

1. Precedence now matches: the writer prefers cadence256 over stroke_rate, the same order as resolve_stroke_rate. Record(stroke_rate=30, cadence256=28.5) writes 28.5 and validate() checks timing against 28.5. I kept both attributes rather than collapsing to one, to avoid changing the public model; happy to do that instead if you prefer. Test added in test_roundtrip.py.
2. The 10–100 spm typical-range warning now checks the resolved stroke rate (the dead _TYPICAL entry is removed). Tests cover 5 and 120 spm warning, and 24 spm not warning.

🤖 Generated with Claude Code