Fix: Stop recurring integration loops on repeated failures + surface unhealthy installations

[matoszz]

• Recurring integration syncs now stop after repeated consecutive failures instead of retrying forever; the integration installation is marked unhealthy and the owning organization gets a notification telling them to reconnect, so customers see actionable status rather than a silently spinning loop
• Connection problems surface at the earliest possible point (ideally), e.g. when an integration's client cannot be established during seeding or a sync cycle, the installation is immediately flagged as needing reconnection instead of queuing work it can never complete
• Platform-owned (AKA "runtime integrations") scheduled sweeps such as billing reminders and organization cleanup are exempt from the failure budget, so they back off through an outage and recover on their own rather than dying silently until the next deploy
• Loop management is now (ideally) self-healing so each operation collapses to exactly one recurring loop, with duplicates purged and missing loops reseeded while a healthy loop keeps its adaptive schedule state
• A shared, mode-switchable test integration replaces bespoke per-suite fixtures, and new end-to-end tests verify the full connect, sync, fail, and reconnect journey against the real queue-backed runtime

[sonarqubecloud]

Quality Gate passed

Issues
2 New issues
0 Accepted issues

Measures
0 Security Hotspots
23.7% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud