fix(framework): Interrupt Fleet retries during teardown

[tanertopal]

What changed

• disable the Fleet retry coordinator before node teardown RPCs
• wake an active heartbeat backoff instead of only lowering its future attempt limit
• verify authenticated connection cleanup invokes retry cancellation

Why

A Fleet heartbeat already waiting in backoff could survive SIGTERM cleanup and keep connection teardown blocked. Explicit cancellation wakes that wait before deactivation and unregister calls run.

Stack

Umbrella/reference: #7895 — Prevent task executor shutdown deadlock

• Previous: #7921 — Preserve output before runtime cleanup
• This PR: fix(framework): Interrupt Fleet retries during teardown #7922 — Interrupt Fleet retries during teardown
• Next: End of stack

Merge the stack in order. After the previous PR merges, retarget this PR to main.

Validation

• pytest py/flwr/client/grpc_rere_client/node_auth_client_interceptor_test.py
• Ruff and Black on the changed connection modules
• git diff --check