Retry fenced workflow task heartbeats after backend loss

[rmcdaniel]

Problem

Server issue durable-workflow/server#137 exposed an unhandled workflow-task heartbeat backend failure during a MySQL interruption. The matching Server change returns a fenced backend_unavailable 503.

Change

• Recognize that typed worker heartbeat response at both root and path-prefixed managed runtime URLs.
• Retry only when the response repeats the submitted task ID, lease owner, attempt, worker identity, and valid retry metadata.
• Fail closed for a mismatched or malformed response.

Verification

• test_storage_admission.py: 94 tests passed in Python 3.12.
• Ruff and MyPy passed for the changed code.

Published SDK and Server artifacts still need the MySQL interruption retest before issue #137 can qualify.