Remove Kinesis producer's internal TTL by default

[msg555]

By default the Kinesis producer implementation will time out records that cannot be delivered within 30 seconds. This interferes with Maxwell's back-pressure implementation as more than 30 seconds of records may buffer before Maxwell begins to throttle itself causing the whole daemon to exit.

I initially tried setting the Ttl to the max value mentioned in the Kinesis producer documentation, unfortunately this isn't actually supported by the Kinesis producer since it appears to cause overflow in the backend and makes the records timeout immediately. Opted for just changing the Ttl to 1 hour instead which I imagine is a reasonable compromise that let's Maxwell's back pressure do its thing while still surfacing an error if records long term continue to be undeliverable.

Tested the change and can confirm I'm not seeing failures in the timeout failures anymore / back pressure is working properly.

Reference

See https://javadoc.io/doc/com.amazonaws/amazon-kinesis-producer/0.14.0/com/amazonaws/services/kinesis/producer/KinesisProducerConfiguration.html#getRecordTtl--

[osheroff]

fine by me, lmk when you want it merged

[msg555]

Should be ready to go. Happy to make any changes you want though.

I've resolved my issue by passing a custom kinesis producer configuration now, so no urgency on my end