Make heartbeat retry metrics test deterministic

[yuandrew]

What was changed

Make heartbeat retry metrics test deterministic by adding a interceptor that waits for activity retries to complete before completing the activity. The exact retry and failure assertions have been preserved.

Supercedes #2759

Why?

Test flake

Checklist

1. Closes

2. How was this tested:

3. Any docs updates needed?

---

Note

Low Risk
Integration test-only change; no production SDK behavior is modified.

Overview
Makes TestLongRunningActivityWithHBAndGrpcRetries deterministic by controlling when the activity finishes relative to heartbeat gRPC retries and metrics.

The test stops the suite worker and dials a dedicated client with a unary interceptor on RecordActivityTaskHeartbeat that signals completion (including retries) on a buffered channel. A custom LongRunningHeartbeat activity sends a fixed number of heartbeats and blocks on that channel after each RecordHeartbeat, so no further heartbeats run while a batched RPC is outstanding and the workflow cannot complete until retries and their metrics have finished. Traffic-controller failures and metric assertions are unchanged, with operation names and expected retry counts pulled into named constants (heartbeatOperation, retryCount, etc.).

^{Reviewed by Cursor Bugbot for commit 04252b4. Bugbot is set up for automated code reviews on this repo. Configure here.}