Add health fault lifecycle tracking and InfluxDB export

[alaurenzi]

What changed

• add aggregator-owned fault lifecycle tracking for standardized health reports
• define a string-native XBot2/ROS diagnostics contract for nodes ending in /health
• track raises and clears by (hw_id, node, fault_report)
• preserve hardware fault state when the diagnostics source becomes stale
• write Grafana-friendly health snapshots and transition-only fault_event points to InfluxDB
• use standardized fault_report values as InfluxDB tags
• add focused lifecycle, contract, ordering, timestamp, and Influx schema tests

XBot2 publisher contract

Only diagnostic names whose final path segment is health participate in fault lifecycle tracking.

A valid health status contains exactly one decimal string count and repeated string reports:

name: /xbot/joint/knee_pitch_1/health
hardware_id: knee_pitch_1
level: 2
message: Drive faults active
values:
  - key: fault_count
    value: "2"
  - key: fault_report
    value: motor over temperature
  - key: fault_report
    value: encoder signal lost

The reports are the complete active set, not deltas. fault_count must equal the number of unique reports.

• no reports + level 0: healthy and clears previous reports
• reports + level 1: warning reports active
• reports + level 2: error reports active
• level 3: source stale; no hardware raise or clear

Malformed or inconsistent messages are ignored for lifecycle purposes rather than clearing existing state.

Publish immediately on set/severity/summary changes and also publish a periodic heartbeat; 1 Hz is the documented default.

InfluxDB schema

health

Written for every health publication, including heartbeats, yielding one latest row per health source for Grafana.

Tags: hw_id, path, name, component.

Fields include level, message, fault_count, active_fault_reports, and—after the first transition—last_fault_report, last_fault_active, last_fault_level, last_raised_ns, and last_cleared_ns when available.

fault_event

Written only on raise/clear transitions.

Tags: hw_id, path, name, component, fault_report, transition.

Fields: active, level, message, occurrence_count, first_raised_ns, last_raised_ns, and last_cleared_ns when available.

fault_report is intentionally a tag because reports are standardized and bounded, with cardinality equivalent to standardized fault codes while remaining directly readable in Grafana.

Both measurements use the source timestamp when valid, with wall-clock fallback.

Documentation and validation

The complete convention and storage schema are documented in docs/fault_health_contract.md.

Tests cover:

• single and simultaneous reports
• duplicate heartbeat suppression for events
• raise, clear, and report replacement
• stale behavior
• /health namespace enforcement
• count and level validation
• transition-before-snapshot ordering
• health snapshot fields
• fault_event report and transition tags
• source timestamp usage

The branch is currently ten commits ahead of master, with six changed files.