Fix Kafka producer queue full bug

[sotojn]

This PR makes the following changes:

• Adds new operations config to the kafka_sender_api called max_buffer_kbytes_size that allows the user to set the total message size sum in kilobytes allowed on the producer queue.
• Adds new logic to the kafka producer that will prevent an error in the case that the queue byte size is reached.
• Fixes false information in the kafka_sender_api to be correct
• Updates docs
• Adds logging to producer flush() actions
• Bumps kafka-assets from v5.1.0 to v5.2.0

Ref to issue #755

[sotojn]

I will be adding debug logging for the producer flush() call in the morning. Moving back to draft until then.

[godber]

On Friday, when reviewing this I remembered there are TWO reasons I wanted you to start test jobs creating these conditions. FIRST to have evidence that the conditions are triggering as needed. SECOND so you can read the logs in the context that they are being printed and assess whether the wording is as effective ... in context of other logs ... as they can be.

Your log messaging is improved for sure, but take a look at worker logs on your test job, and think about what can be improved. There's one word you can add to the message that would improve it immensely. This word is not obvious when you're implementing it because the context of your change is super obvious, but in the mess of other log messages, from the perspective of the user/operator it is less obvious. Technically the word is in the output, but not in the log message itself.

Respond as a comment here, then we'll wrap up this PR.