Option to use the KPL

[rgarcia]

The KPL brings many benefits, namely aggregation, which would significantly reduce the number of records per second generated by the agent (#5, #57, #31).

If this is something you're open to, I would be happy to try to implement this with some guidance on what the feature should look like.

[leandro-rouberte]

I'm also interested to see Kinesis agent using KPL. I'm developing a pipeline and I'm having to provision more shards in Kinesis stream that is actually needed (considering bytes volume), because of the lack of aggregation functionality.