fix(operator): recreate deleted MCPServer StatefulSet

[RaviTharuma]

Cross-links

• Issue: bug(operator): deleted MCPServer StatefulSet is not recreated #6343
• Maintainer direction (watch/recreate STS + document dual-pod model): bug(operator): deleted MCPServer StatefulSet is not recreated #6343 (comment)
• PR announcement: bug(operator): deleted MCPServer StatefulSet is not recreated #6343 (comment)

Summary

Deleting the runner-created MCPServer StatefulSet left the proxy Deployment
up and Ready=true. Clients hit a dead backend. The operator created the
Deployment but did not watch/adopt the STS, so orphan-cleanup scripts treated
it as leftover.

• Watch ToolHive StatefulSets and adopt a controller owner-ref on the
  MCPServer
• If the STS is missing and the proxy is already Available, bounce the proxy
  (existing restarted-at annotation, 2m cooldown) so the runner re-applies
• If the proxy is not yet Available, only requeue (do not interrupt first
  boot)
• Status: proxy running + STS missing → Pending, Ready=False
• Document the dual Deployment + StatefulSet model

Fixes #6343

Type of change

• Bug fix
• New feature
• Refactoring (no behavior change)
• Dependency update
• Documentation
• Other (describe):

Test plan

• go test -ldflags=-extldflags=-Wl,-w ./cmd/thv-operator/controllers/ -run 'TestEnsureWorkloadStatefulSet|TestMapStatefulSetToMCPServer|TestRecentlyBounced'
• CI unit / operator tests
• After merge: deleting the workload STS should bounce the proxy, recreate
  the STS, and keep MCPServer Pending until it is back

API Compatibility

• This PR does not break the v1beta1 API, OR the api-break-allowed label is applied and the migration guidance is described above.

No CRD schema change. RBAC already includes StatefulSets.

Does this introduce a user-facing change?

Yes. A deleted MCPServer workload StatefulSet is recreated on the next
reconcile, and Ready is no longer claimed on a proxy-only stack. See
docs/operator/mcpserver-workloads.md.

Special notes for reviewers

The operator still does not author the STS spec — the proxy-runner does via
SSA. Recreate is "bounce the runner so it applies again," which matches how
the STS is born. Cooldown avoids a restart loop while the runner is still
starting.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[RaviTharuma]

Implements #6343 per maintainer direction.

Cross-links:

• Issue: bug(operator): deleted MCPServer StatefulSet is not recreated #6343
• Maintainer comment: bug(operator): deleted MCPServer StatefulSet is not recreated #6343 (comment)
• Issue PR note: bug(operator): deleted MCPServer StatefulSet is not recreated #6343 (comment)

[reyortiz3]

@RaviTharuma mind fixing the lint error?

[codecov]

Codecov Report

❌ Patch coverage is 61.00000% with 39 lines in your changes missing coverage. Please review.
✅ Project coverage is 77.67%. Comparing base (406fd8d) to head (a547117).
⚠️ Report is 21 commits behind head on main.

Files with missing lines	Patch %	Lines
.../thv-operator/controllers/mcpserver_statefulset.go	64.70%	24 Missing ⚠️
...d/thv-operator/controllers/mcpserver_controller.go	53.12%	15 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6379      +/-   ##
==========================================
+ Coverage   72.98%   77.67%   +4.69%     
==========================================
  Files         744      751       +7     
  Lines       78611    72694    -5917     
==========================================
- Hits        57374    56467     -907     
+ Misses      17236    16222    -1014     
+ Partials     4001        5    -3996     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.