implement jlap format for incremental repodata patches

[wolfv]

[wolfv]

I think the main difficulty (similar to zchunk) will be to have a proper "state-machine" for the different error cases. It is complicated by the fact that the idea of the jlap format is to be able to continue writing to the jlap file from a given offset.

• no repodata json at all locally -> download repodata.json
• repodata.json locally -> check if repodata.jlap exists on server
• jlap does not exist on server -> update json
• jlap does exist, but our local jlap / offset does not match server -> redownload entire jlap file
• jlap exists, but -hash of repodata.json is not found in jlap file -> redownload repodata.json (& remove jlap file?)
• jlap exists, from-hash exists, but at some point the hashes diverge -> redownload repodata.json (incorrect jlap file)
• parse error of jlap file (e.g. missing required fields)

Maybe there are more cases but these are the ones that come to my mind right now. Probably we should have a dedicated error class for each case.

[dhirschfeld]

TIL the jlap format from a post on discourse:

• https://conda.discourse.group/t/how-we-reduced-condas-index-fetch-bandwidth-by-99/257

It sounds very interesting so I'll make sure to follow progress here...

[jonashaag]

What’s the status of this PR?