cloud_storage: The remote partition reader can get stuck

[Lazin]

Version & Environment

Redpanda version: (use rpk version): dev

What went wrong?

In case if LSO is low (inside the TS range) the read_some method can get stuck by returning an empty record batch set.
The internal read loop of the reader goes into an infinite loop in this case. In order for the problem to reproduce we need to have some transactional batches in the log. This batches has to be skipped to proceed but we fail to skip them correctly because our LSO does not allow this.

It's not clear why this could happen because when we're reading near the LSO we're supposed to read from the local storage. Probably, we're start reading from the offset which is only available in the cloud storage.

What should have happened instead?

The reader shouldn't get stuck.

How to reproduce the issue?

N/A

Additional information

N/A

JIRA Link: CORE-2321