[vuongvm812]

Summary

• Part of a stack: run-metrics-garmin (bottom) adds the charts; this PR (top) supplies Huawei Band 10 data. They render from the same new kinds once both merge. Not to be merged on its own.
• Adds DeviceMetricKind entries CADENCE("cadence"), POWER_W("power_w"), ALTITUDE_M("altitude_m") (Kotlin only, no migration). Swift is untouched: it has no Huawei ingest, and the Garmin PR adds the Swift entries.
• Carries per-sample step rate / cadence, altitude and power from the band's 17/0A workout pages into deviceMetricSample through the existing insert. Absent, sentinel (0xFF / 0xFFFF / raw -1) and 0 values are not stored; samples outside the workout are dropped.
• Gadgetbridge import reads STEP_RATE, CADENCE, ALTITUDE, CYCLING_POWER when the columns exist.
• One always-on strap-log line per synced or imported workout with the usable sample counts for HR / cadence / altitude / power.

Changes

Area	Change
data/WhoopRepository.kt	three new kinds appended after SKIN_TEMP_C
ingest/HuaweiWatchIngest.kt	workoutSeries() extracts the rows and counts; fold stores them with source huawei-17-0a (or gadgetbridge)
data/HuaweiHistoryIngest.kt	store(..., log) logs workoutLine per workout
ingest/Gadgetbridge{Importer,Mapping}.kt	extra columns (guarded by the existing column helper) and a log hook, wired in DataSourcesScreen
ATTRIBUTION.md, Tools/parity_*.json	facts-only attribution; platform_specific dispositions ("Huawei ingest exists only on Android")
tests	HuaweiWorkoutSeriesTest, extended GadgetbridgeImporterTest, kind-list pin

Decisions, from Gadgetbridge's HuaweiWorkoutGbParser.buildActivityPoint: cadence is steps/min, taken from the step rate (field 2) for walking (types 2, 13) and otherwise from the inner-block cadence (bit 0), each falling back to the other. Altitude is stored in tenths of a metre (raw / 10). The step rate is read as an unsigned byte (& 0xFF), with raw -1 meaning absent. Power is raw watts.

How to Test

1. cd android && ./gradlew testFullDebugUnitTest (create app/build/roomSchemaOracle if Gradle complains).
2. python3 Tools/parity_ledger.py --base origin/main, doc_comment_lint.py, i18n_audit.py --ci origin/main.
3. On a device: sync a Band 10 workout (or import a Gadgetbridge export) and look for Huawei: workout <ts> type=N usable samples hr=.. cadence=.. altitude=.. power=.. in the strap log.

Verification

• Android unit suite: 7303 tests; the only failures are the known main ones (HuaweiSyncedEngineTest, RecoveryDriversTest x2).
• Parity ledger clean. parity_ratchet.py still prints "base authority cannot be reproduced with the current derivation", which comes from main's base, not this change.
• Tools unittest: 4 test_check_source_references failures from macOS /var vs /private/var paths, unrelated.
• Not tested on hardware: decoding is untested on a real Huawei Band 10 (and no Garmin file for the sibling PR). The Gadgetbridge export at hand has all these columns at 0, so only a synthetic fixture was exercised.

Known Limitations / Follow-ups

• A 0 count in the log can mean the band or sport does not provide the series; the line states what was decoded, not what the band supports.
• Altitude of exactly 0.0 m is skipped with the other zeros, although sea level is a real altitude.
• GPS-file altitude (GpsPoint.altitudeM) is still not stored.
• The Gadgetbridge columns' signedness is inferred from its & 0xFF use; both readings map to the same value.
• A restore of an Android .noopbak on Apple relies on Swift tolerating unknown kinds (as for skin_temp_c).
• Reviews: code-review skill and Codex (Codex found nothing and could not run tests).