[DynamoDB] Avoid OOMing from snapshotting

[Tang8330]

Right now, we're reading the entire json.gz file and then writing it.

DynamoDB chunks the files for 1.5M rows per file by default. However, if the row is really large - we're susceptible to OOM errors

[Tang8330]

Hmm, closing this. I think we need to throttle StreamJsonGzipFile, and this doesn't do that exactly.

[Tang8330]

Actually, this is fine. We are using an unbuffered channel.

Unbuffered channels provide synchronous communication, meaning the sender will block until a receiver is ready to receive the value, and vice versa.