Implement exclude filters

[muturgan]

Closes #5

[vbrandl]

Using multiple cache files, for each set of excluded files is way simpler then changing the cache format as I had it in mind. Good idea.

I will test this soon.

Thinking about it: having multiple cache files breaks the delete_repo_and_cache function. To fix this, we would have to use something like glob, but deleting cache/github.com/username/reponame*.json would also delete caches for every other repository of username which name starts with reponame.

I think the cleanest and correct way would be to add another layer into the cache format and make it a HashMap<Branch, HashMap<Excludes, CacheEntry>>. And while at it, there should be some kind of locking mechanism when accessing the cache. Currently two parallel request might cause conflicting cache writes, causing a corrupt cache.

[muturgan]

Well, need think about it

[muturgan]

I think delete all files by pattern is what we need. If a hoc score has changed on all repository - this has changed (most likely) for a some subset of files.
But as you say branch is important, need add it to a cache file path.

[muturgan]

May access to a cache via rwlock

[muturgan]

@vbrandl do you know a cache size on prod? Can we store it in memory?

[vbrandl]

Using an in-memory cache would also be fine, I think. Could you first describe, how you plan to implement said cache?