Orphaned availabilities

[AuHau]

I am not sure if it is a problem or not, but I have realised that Availabilities will be most probably never "used-up" as most probably there will be some capacity left in them (even very little) for which there won't be the proper Request's size (and other parameters) fit.

Is it a problem? WDYT? @markspanbroek @emizzle

[markspanbroek]

The small leftovers should probably be merged with other compatible availabilities (e.g. with the one it originally split off from when it returns (#410))

[emizzle]

This is a good point. I think we could have availabilities under a certain size (cli param-defined perhaps) try to merge with compatible exisiting availabilities (as @markspanbroek mentioned above) We would also need to consider attempting a merge when new availabilities are added, as there may still be availability dust floating around. As part of #455, there will be a signal triggered when availabilities are added, which could be utilised here.

(e.g. with the one it originally split off from when it returns (#410))

I'm not sure there is a "splitting" of availabilities. The current implementation simply reduces the availability's size as blocks are downloaded, then marks it unused after download. Once that availability gets to a certain size, we could then attempt to merge that availability with other availabilities that have looser (or equal) availability requirements.

[AuHau]

This will be mainly solved by switching to "persistent availabilities" by default which is something I will be working on at #467