improvement: Add graceful handling for disk-full conditions

[tphakala]

Summary

When the database disk is full, multiple subsystems fail simultaneously with cascading errors. There is no proactive disk space check or graceful degradation.

Sentry References

• BIRDNET-GO-SW: Alerting save_history fails (5 events)
• BIRDNET-GO-SX: FFmpeg audio export broken pipe / no space left (33 events, escalating)
• BIRDNET-GO-SZ: Save threshold event fails (4 events)
• BIRDNET-GO-T0: Save threshold event fails (1 event)
• BIRDNET-GO-T1: SQL query fails (1 event)
• BIRDNET-GO-T2: Save hourly weather fails (1 event)
• BIRDNET-GO-SY: Commit transaction fails (1 event)

All from the same category: database or disk is full (SQLite sqlite3.Error).

Error Details

[database] database or disk is full
[system-resource] failed to write PCM data to FFmpeg: No space left on device

Impact

When disk fills up, at least 7 different error types fire simultaneously, creating noise. The system continues to attempt writes that will fail.

Fix Suggestions

1. Proactive monitoring: Check available disk space periodically and warn before it's critically low
2. Circuit breaker: When a disk-full error is detected, temporarily pause non-critical writes (weather, alerting history) to reduce error noise
3. Diskmanager integration: The diskmanager should proactively clean up old recordings before disk fills completely
4. Single error reporting: Deduplicate the disk-full condition into a single high-priority alert rather than per-operation errors
5. User notification: Surface a clear UI warning when disk space is critically low

Related Code

• internal/datastore/ — all database write operations
• internal/myaudio/ — FFmpeg audio export
• internal/diskmanager/ — disk space management