fix(server): index & gc queue & llm compatibility (#15528)

fix #15523
fix #15526

#### PR Dependency Tree


* **PR #15528** 👈

This tree was auto-generated by
[Charcoal](https://github.com/danerwilliams/charcoal)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

- **Bug Fixes**
- Improved search generation cleanup, provider error reporting, and
reconciliation reliability.
- Retired search resources are cleaned up safely, including after
credential changes.
- Prompt size checks now ignore tool parameters and provide clearer
errors.
- Reserved documents are protected from accidental cleanup, and
malformed identifiers are rejected.

- **Configuration**
- Managed Copilot profiles require explicit, non-duplicated model
assignments.
  - Improved managed provider profile migration.

- **Performance & Reliability**
- Reduced unnecessary search-history cleanup and adjusted
consistency-check intervals.
- Failed reconciliation jobs stop after one attempt and are removed
automatically.

- **Data Updates**
  - Updated legacy AI session prompt names to current labels.
  - Improved cloud load-balancer health-check configuration.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

.github/helm/separate-config/backend-config.yaml

@@ -1,10 +0,0 @@
-apiVersion: cloud.google.com/v1
-kind: BackendConfig
-metadata:
-  name: "affine-api-backendconfig"
-spec:
-  healthCheck:
-    timeoutSec: 1
-    type: HTTP
-    requestPath: /info
-