Prioritize "Ready" Pods in SandboxClaim adoption logic

[janetkuo]

The SandboxClaim controller currently adopts the oldest pod from the warm pool regardless of its state. If the oldest pod is unhealthy (e.g., CrashLoopBackOff or Pending), the claim adopts it and stalls, defeating the purpose of a warm pool.

In the adoption logic, we should prioritize pod health:

1. Sort pods primarily by Readiness (Running & Ready condition).
2. Use CreationTimestamp (oldest first) only as a tie-breaker.

We could consider supporting different types of adoption strategy in claim, such as AdoptionStrategy: Any | PreferReady | StrictReady.

/kind feature

[peterzhongyi]

I'll send out a fix for this.

[peterzhongyi]

/assign peterzhongyi