feat(log-analysis): add Plane landing analysis

[nflyernz]

Summary

Adds ArduPlane 4.7.x landing analysis to AMC using the existing
log-analysis architecture and flat result model.

The implementation is based on the validated landing behavior from
ArduPilotTools (APT), migrated onto AMC-native:

• LogData
• LogAnalysisContext
• PlaneFlightSegmentDetector
• ParameterHistory
• availability/analysis registry
• flat LogAnalysisResult findings

No Plane-specific backend orchestration or frontend path is introduced.

Scope

Initial support is intentionally limited to ArduPlane 4.7.x.

Required evidence:

• GPS
• LAND
• MODE
• BARO

Optional evidence:

• ARSP
• RFND
• CMD
• MSG

The subsystem also requires at least one operational Plane flight segment.

Landing attempt detection

Supports:

• multiple landing attempts within one flight
• aborted/restarted approaches
• go-arounds without splitting the parent flight
• AUTO landing-stage detection
• GPS-stop termination
• abort-message termination
• mode-exit termination
• disarm termination
• parent flight-segment termination

Across the four established APT regression logs:

• 10 operational flights
• 16 landing attempts
• 16/16 attempt boundaries and associations match the validated APT behavior

Landing evidence

Adds objective flat findings for:

• LAND stage 2 preflare timing/height
• LAND stage 3 flare timing/height
• preflare/flare GPS speed
• ARSP evidence where available
• BARO altitude and preflare sink rate
• RFND stage-point evidence
• event-time landing parameters using ParameterHistory
• firmware landing glide-slope messages
• firmware flare altitude/sink/airspeed/distance
• flare-to-GPS-stop duration

Firmware comparison against the reference logs:

• 16/16 glide-slope events match
• 14/14 firmware flare events match

Mission LAND target evidence

Implements conservative CMD snapshot reconstruction matching APT:

• complete ordered mission snapshots only
• latest complete snapshot applicable at attempt start
• exactly one MAV_CMD_NAV_LAND
• malformed/incomplete/ambiguous mission state rejected
• independently computed GPS-stop to mission LAND-target distance

Reference comparison:

• 16/16 mission LAND-target selections match APT
• 3/3 computed target distances match APT

Rangefinder lifecycle evidence

Adds the validated APT RFND lifecycle evidence:

• first nonzero sample
• first in-range sample
• continuous acquisition
• disengagement count
• last disengagement

All nine RFND lifecycle fields match across all 16 reference landing attempts.

RFND remains optional.

AMC-specific correctness

Landing GPS consumers use AMC's existing active-receiver convention:

• if GPS.U exists, only U == 1 is used
• otherwise all GPS records are used

Event-time landing parameter findings also reject non-finite values rather
than emitting NaN/Inf evidence.

Architecture

Follows ARCHITECTURE_log_analysis.md and the documented
"Extending the Log Analysis System" path.

The PR:

• uses existing LogData and LogAnalysisContext
• uses the existing Plane flight segmentation infrastructure
• uses shared ParameterHistory
• registers the Plane landing subsystem through the existing registry
• emits existing flat analysis results
• makes no extraction/backend/frontend architecture changes

Deliberately out of scope

Not included in this PR:

• touchdown detection
• post-attempt rollout evidence
• firmware Distance from LAND point=...
• widening completed FlightSegments
• runway/cross-track/along-track geometry
• qualitative landing scoring
• parameter recommendations
• Plane landing configuration-step/frontend changes
• hierarchical per-flight/per-attempt results
• firmware support outside ArduPlane 4.7.x

Validation

Final applicable non-SITL AMC test run:

• 4901 passed
• 55 skipped
• 4 xfailed

Additional validation:

• focused Plane landing tests pass
• APT landing regression passes
• APT event/timeline regression passes
• Ruff lint passes
• Ruff format check passes
• changed-file mypy passes
• changed-file pylint 10.00/10
• git diff --check passes

[coveralls]

Coverage Report for CI Build 33926127272

Coverage increased (+0.1%) to 89.446%

Details

• Coverage increased (+0.1%) from the base build.
• Patch coverage: No coverable lines changed in this PR.
• 12 coverage regressions across 1 file.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

12 previously-covered lines in 1 file lost coverage.

File	Lines Losing Coverage	Coverage
log_analysis/data_model_log_analysis.py	12	89.34%

---

Coverage Stats

Relevant Lines:	19717
Covered Lines:	17636
Line Coverage:	89.45%
Relevant Branches:	5942
Covered Branches:	4911
Branch Coverage:	82.65%
Branches in Coverage %:	No
Coverage Strength:	2.66 hits per line

---

💛 - Coveralls