Batch processing of notifications

[c0c0n3]

PR #191 added support to store multiple entities in notification messages---btw, kudos to @ohylli. But we still need to think about how to handle:

1. Partial errors. What to do if we encounter an error for some of the entities in the batch? e.g. what if 2 out of 20 entities fail validation? should we reject the whole batch (probably not a good idea!) or try saving the 18 that passed validation? What would the HTTP response look like in this case? How do we accumulate errors?
2. Message size. What to do with large batches---e.g. 100s of entities? Ideally we should process them in constant space to avoid eating up all available RAM. This applies both to the received HTTP message and the DB call we issue.

For now this is just a reminder that at some point in time we'll have to consider the points above. Also, we should probably split (1) and (2) in two separate PRs.

Initial thoughts about a possible way to implement

1. Leverage monoids/monads/monoidal functors as suggested on stackoverflow
2. Use stream processing techniques to process a batch of batches of configurable size.

[github-actions]

Stale issue message