Kafka consumer reliability/performance improvements

[noggi]

• Reliability: check/retry commit failures. By default, commit is asynchronous, and no error checking/logging is performed. This PR converts the commit call to a synchronous call, checks if any partition failed to commit, logs the errors, and retries failed commits.

• Performance: originally, actions consumer commits offset after passing every single message to the user. Commit is very expensive by itself, and it causes a network round-trip, so processing of a heavily backed up topic may take a very long time. This changes the behavior of the consumer such that it only commits after messages that the user cares about, and all other messages are committed periodically, according to the standard auto-commit configuration.

@jjoyce0510

[github-actions[bot]]

Unit Test Results (build & test)

63 tests  ±0   63 :heavy_check_mark: ±0   3s :stopwatch: ±0s   1 suites ±0     0 :zzz: ±0    1 files   ±0     0 :x: ±0 

Results for commit 3845f7e3. ± Comparison against base commit 01661f0e.

:recycle: This comment has been updated with latest results.