fix(receiver): harden callback draining during shutdown

[GefMar]

• keep callback tasks owned until cleanup completes
• prevent AnyIO level-cancellation busy loops
• preserve interruptible unbounded graceful waits
• release execution capacity exactly once
• clarify wait-tasks-timeout semantics
• add deterministic lifecycle and cancellation tests

Problem

Receiver callbacks could become orphaned or remain incompletely drained during worker shutdown.

The issue occurs when a message callback is still running while the receiver is being cancelled, especially when:

• --wait-tasks-timeout expires;
• the worker receives SIGINT or SIGTERM;
• cancellation is delivered repeatedly;
• an AnyIO cancel scope continues applying level cancellation during cleanup.

In these cases, cancellation could interrupt the callback-draining path itself. This could leave callback work without a clear owner, skip cleanup, or release prefetch capacity incorrectly. Repeated AnyIO cancellation could also cause a busy loop while the receiver attempted to finish cancellation cleanup.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.92%. Comparing base (8a1b04d) to head (4ffc46f).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #644      +/-   ##
==========================================
+ Coverage   85.37%   85.92%   +0.55%     
==========================================
  Files          68       68              
  Lines        2660     2743      +83     
==========================================
+ Hits         2271     2357      +86     
+ Misses        389      386       -3     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.