HTTP requests are the bottleneck of the resolve time.

Hashes are stable enough we can cache them when found.

Plan is:

• Provide a small cache
  • plain files, sqlite3 or b-tree à la db5?
  • if 34835825f73 is cached, do we use it for 34835825 too?
  • $HOME/.cache/resolve-hash could be an ideal location
• Code integration
  • create a new class CacheHashSearch
  • define it as the first entry of get_search_classes()
  • check after a search the source of resolution, if it's not from CacheHashSearch, cache it