Asset modified cleanup rfc document

[JackLewis-digirati]

This PR adds an RFC discussing #430

[donaldgray]

It's not explicitly mentioned but cleanup logic will need to take into account CustomerOriginStrategy, policy for image (use-original/default) and other available channels (i.e. it can't just look at the delivery-channels individually).

A thought when reading: How would this work during period of mass reingestion? ie if 10K..100K..1M..10M items are queued - items added last in the queue could take hours/days to be picked up by Engine. The max value for delay_seconds is 15mins so everything will be first picked up pretty quickly - assuming Engine is still in flight they wouldn't be acted on. As mentioned we can use visibility_timeout (max 12 hours) to ensure it's not picked up again for a while but there are limits on the number of in-flight messages. The limit is large and we're using long-polling which means we wouldn't receive an error but would receive empty responses until there are less messages in-flight. Do we need to bear that in mind for this doc? Even if it's a backup solution to manually disable asset-cleanup if we know a mass ingest will happen.

General comment, there's inconsistent capitalisation + use of full stops