RcloneListDirIter: unify list chunk size for agent and backend

[Michal-Leszczynski]

Basically all backends chunk dir entries when listing. E.g. s3 has the MaxKeys param which is set to 1000 by default.
We also have our own chunking mechanism implemented:

scylla-manager/pkg/rclone/rcserver/list.go

Lines 12 to 17 in b97d1cf

	type listJSONEncoder struct {
	enc *jsonEncoder
	buf []*rcops.ListJSONItem
	maxItems int
	started bool
	}

In the past we used the same chunk size in the backend queries and in our own chunking mechanism, which made sense.
Starting from commit 43230ec, we don't set the backend chunk size anymore. This means that we use the default chunk size for the backend (e.g. 1000 for s3), and a different value for our own chunking mechanism:

scylla-manager/pkg/rclone/rcserver/list.go

Line 10 in b97d1cf

const defaultListEncoderMaxItems = 200

I believe that it was mistakenly forgotten during refactor and that we should align those chunk sizes.
Let's go with the intentional 200.

[Michal-Leszczynski]

Actually, this is configurable via scylla-manager-agent.yaml config file:

s3:
  list_chunk: "200"

So let's not change the default yet, as it might affect other clients.
We should just document that changing this value might help with the rclone list dir timeout errors.