Hash generation batch can skip items, loop infinitely

Problem/Motivation

The generation batch uses a numerical position offset within the result set, which means that if one or more already-processed/hashed entities happen to be deleted while the batch is running, then that number of items could be skipped by the batch process.

Steps to reproduce

Haven't actually run into this, more planning to enable and run the batch in an active site. Hypothetically, should be able to trigger with:

1. Create two managed file entities
2. Enable/configure filehash
3. Introduce a pause at the end of the batch process callback
4. Execute the hash generation batch
5. While in the pause, after processing the first entity, delete the first entity

The second entity would not presently get processed.
The actual failure mode appears to be a non-terminating loop: The DB query should just return an empty result set which we will iterate over forever while returning that we are not $context['finished'] because we do not further increment $context['results']['processed'] and $context['sandbox']['count'] is stale (as is the $context['results']['processed'], since it is counting that it processed something that is no-longer in the system).

Proposed resolution

Track and use the last fid to offset in the query. The query is already ordered by fid.
Exit the batch when the query returns an empty result set.

Remaining tasks

User interface changes

API changes

Data model changes