[feature] Add GRPC Retry Logic to Metrics Collector

/kind feature

**Describe the solution you'd like**
[A clear and concise description of what you want to happen.]

Metrics collector collects the metrics from main container and report them to manager via GRPC. We should have sort of retry logic for GRPC calls.

**Anything else you would like to add:**
[Miscellaneous information that will assist in solving the issue.]

Start by locating the metrics collector entry point and the existing gRPC calls that report metrics to the manager. Review any nearby tests or error-handling conventions, then clarify retry conditions, limits, and backoff before implementing; done means transient gRPC failures are retried without disrupting metric collection.

Written by the indexing model from the issue text.