feat: add Prometheus metrics exporter endpoint (GET /metrics)

[bolausson]

Summary

• This is implemented with the help of an LLM!
• Since I don't have the necessary hardware yet, I haven't been able to test the metrics for plausibility
• Add GET /metrics endpoint exporting ~60 metrics in Prometheus text exposition format
• Covers all sensor readings, derived values (dew point, absolute humidity, mold risk), DAC/fan control state, network status, and ESP32 system diagnostics (heap, PSRAM, chip temperature, flash sizes, FreeRTOS task count)
• Invalid/unavailable sensors are omitted rather than zeroed, which is the correct Prometheus semantic for absent data
• Uses chunked HTTP transfer to keep peak memory usage at ~1.5 KB
• Flash/sketch size metrics are cached at init to avoid repeated SPI flash reads on every scrape

Files Changed

• New: src/web/PrometheusExporter.h, src/web/PrometheusExporter.cpp — exporter implementation
• New: test/test_prometheus_exporter/test_prometheus_exporter.cpp — 10 native unit tests
• New: docs/prometheus/README.md — full metric reference, scrape config examples, PromQL queries
• Modified: src/modules/NetworkManager.cpp — route registration and init
• Modified: platformio.ini — added exporter to native test build filter
• Modified: README.md — added Prometheus section with quick-start config

Testing

• Firmware compiles cleanly with zero warnings (pio run -e project_aura)
• All 35 native unit tests pass (pio test -e native_test), including 10 new tests covering sensor metric serialization, invalid sensor omission, derived metric computation, pressure trend output, and Prometheus format correctness (HELP/TYPE/value ordering)
• On-device testing was not performed as hardware is not yet available. The endpoint follows the exact same data access patterns as the existing /api/state handler and should be validated with curl http://aura.local/metrics and promtool check metrics once hardware is available

Checklist

• I have read CONTRIBUTING.md and agree to the CLA.
• I tested my changes or explained why tests are not applicable.
• I updated documentation if needed.

[21cncstudio]

Thanks a lot for the contribution and for the detailed implementation.
The Prometheus /metrics work looks solid and is very much appreciated.

For now, I’m focusing on core firmware stability and release scope, so this is a bit early for the current roadmap. I’d like to revisit this in a future phase when we expand observability features.

I reviewed the changes locally and the idea is great. It’s just a timing/prioritization decision for this release cycle.

Thanks again for the effort and for putting this together.

[bolausson]

That is fine.