Heuristic when to fetch full repodata

[pavelzw]

For queries like this:

rattler search '* >=3'

it doesn't make sense to use sharded repodata as this would result in more than 10k fetches.
Maybe if >x% of packages in repodata names (or an absolute number(?)) would be requested by a rattler repodata gateway query, we should fall back to full download instead?

WDYT @baszalmstra

[baszalmstra]

Yeah that makes sense, the same goes for supporting queries like, which packages depend on x.