Kafka Connect isn’t a Snowplow product so I’m not sure we can give you much assistance with it. Confluence have extensive documentation on it, so I imagine one of their forums might be more a fruitful avenue for questions specific to Kafka Connect.
On the overall architecture, I will mention that we don’t currently support a Kafka → S3 → Redshift architecture. I think what you propose is workable, however I wouldn’t necessarily expect it to ‘just work’ out of the box. The format and filestructure in S3 needs to match that produced by the S3 loader (which loads from Kinesis to S3), so more steps may be needed to get the data into shape before the loader would work.
The easier way to get data into Redshift of course is to go with the Kinesis-based pipeline, but I do understand that there are reasons one would prefer Kafka.