cache mock-up

[fluvf]

Accidentally based on #42
One monolithic commit as this is a mock-up
Notable changes needed to make this work:

• Fallback IDs are now searched for in reverse
  Simplifies logic and allows subsequent non cached find_entry_paths to overwrite any set aliases (Not actually strictly needed?)
  It is, if cached find finds a duplicate entry that's of lower priority than one found during a full search, the higher priority entry is skipped No it's not, that entry alias would be unset
  Nevertheless, I will create a separate pull request with only this change later
• find_entry_paths call and $FALLBACK_ENTRY_IDS was moved inside find_entry
  The function needs to see the masked find function set inside find_cached_entry, and the variable needs to be reset between calls
  There's actually no need for it to be a global any more

I don't know if I like sourceing the cache file, might change that to read loop when cleaning up, if you're happy with the implementation

[fluvf]

Thoughts? Would you be happy with this implementation? I'd have time to clean this up today if you are

[Vladimir-csp]

Thank you. I will dig deeper in it a bit later. What I can see right away is that entry read is still needed.

With current cache implementation, exec line is read directly from the cache.

[fluvf]

Yes, this isn't trying to be faster than the current implementation. Some quick tests prove that it's not.
But it is less code to maintain, and the cache should be much harder to invalidate