fix: Handle dilutions, replicate measurements, and mmol/l unit in Roche Cedex Bio HT

[nathan-stender]

Summary

A V6/V7 Roche Cedex Bio HT export failed to parse. Three distinct issues surfaced, each fixed here:

1. Dilutions treated as duplicates. The same analyte can be measured at multiple dilutions within one sample (e.g. IgG human undiluted 3.933 and at 1/5 → 6.174, a minute apart). The dilution factor lived in the previously-discarded col8. It's now named dilution factor, threaded through RawMeasurement, and included in the analyte grouping key so the two measurements no longer collide. The factor is also preserved in the analyte custom information document.

2. Genuine replicate measurements raised an error. The same analyte can be re-measured (a second round) inside the 1-hour grouping window (e.g. glucose 15.17 then 15.43 ~43 min later). Instead of raising Duplicate measurement, we now split the later measurement into a new group keyed by its own time — realizing the improvement the previous code's comment explicitly anticipated.

3. Lowercase mmol/l unit. A handful of rows report the unit as mmol/l; it's now normalized to mmol/L via UNIT_CONVERSIONS.

Changes

• constants.py: rename col8 → dilution factor in DATA_HEADER_V6_V7; add mmol/l → mmol/L to UNIT_CONVERSIONS.
• roche_cedex_bioht_structure.py: add dilution_factor to RawMeasurement, include it in the grouping key and custom info; split on duplicate instead of raising (removes now-unused AllotropyParserError).
• Updated affected unit tests; added test_create_measurements_same_analyte_different_dilution and reworked the duplicate test to assert the split behavior.
• Added trimmed regression file roche_cedex_bioht_v7_dilution_and_replicates.txt (+ expected JSON) covering all three cases.

Testing

• Full 2.4MB source export now parses (5864 groups).
• All 22 roche_cedex_bioht tests pass; hatch run lint (ruff, black, mypy) clean.

🤖 Generated with Claude Code