[giscebot]

Summary

• measure available capacity on the filesystem that contains the SQLite database
• alert when either configurable threshold is crossed (default: 5 GiB or 10% available), and fail closed when capacity cannot be measured
• propagate thresholds through the monitor alert service and document the systemd environment settings

Incident evidence

At 2026-11-11 17:12:28 CET, heartbeat commits failed with database or disk is full; all executor workers then hit disk I/O error and the executor restarted. A read-only check with this branch currently reports only 2.6 GiB / 0.9% available on the database filesystem, so host cleanup remains an urgent operational action. This PR adds the missing early warning; it does not reclaim disk space.

Validation

• pytest -q -p no:cacheprovider — 720 passed
• focused monitor tests — 60 passed
• live read-only monitor check — emitted monitor.database_filesystem_low with 2.6 GiB / 0.9% available

Addresses #343
Addresses #344
Addresses #345
Addresses #346