Raise timeout for unfinished producer futures at deadline

[AryanRG]

KafkaProducer.send(...).get(timeout=0) can currently return None while the record is still pending. A positive timeout can do the same if a batch-split wakeup reaches the deadline before the next loop iteration. An unfinished future should raise KafkaTimeoutError, not appear to have returned a delivery result.

Move the deadline check inside the wait loop so completed success/error results still win, while unfinished calls raise when the budget expires. Add MockBroker regressions for pending sends, eventual success/failure after a timeout, and a controlled rebind at the deadline.

Local validation on macOS / Python 3.13.12:

• Three new cases fail on unchanged master; all five focused cases pass with the fix.
• Producer and future suites: 312 passed with the selector backend, and 312 passed with the asyncio backend.
• Changed production file passes pylint errors-only and git diff --check.

No real broker, full repository suite, other Python versions, or hosted CI was run locally. AI assistance was used to investigate, implement, review, and test this change.