MCPRemoteProxy config-handler errors leave Ready=True while Phase=Failed #6770

Bug description

validateAndHandleConfigs (cmd/thv-operator/controllers/mcpremoteproxy_controller.go:232) runs six config handlers; on error each writes only proxy.Status.Phase = Failed before Status().Update and returns. No Status.Message, no Status.ObservedGeneration, and the Ready condition is never touched — while every other failure path in the same controller writes all three (:86-95, :132-145, :326-336, :373-388). A proxy that was healthy and then loses a referenced MCPToolConfig therefore reports both verdicts at once:

phase: Failed
message: Remote proxy is running            # stale, from :1643
observedGeneration: <last healthy generation>
conditions:
  - type: Ready
    status: "True"                          # stale, from :1644-1649
    reason: DeploymentReady
    message: Deployment is ready and running

kubectl wait --for=condition=Ready and anything keying on Ready still see a healthy proxy while the operator has declared it Failed.

Steps to reproduce

Apply an MCPRemoteProxy with a valid MCPToolConfig reference, wait until it reports Ready, then delete or break that config and re-read the status after the next reconcile.

Expected behavior

Phase: Failed together with a Message naming the failed config, the current ObservedGeneration, and Ready=False — what :86-95 and :326-336 already do.

Actual behavior

Phase: Failed with the previous healthy Message and Ready=True left in place, and ObservedGeneration still on the last healthy generation.

Environment (if relevant)

• ToolHive version: c23718d5 (main)

Additional context

Found while sweeping this controller for the same pattern as #6726, whose fix is in #6765; reyortiz3 asked for this one as a separate issue. I plan to submit a PR mirroring :86-95 on the six paths, with a regression case per handler.