Caching

[seveas]

Currently only blame-tree output (well, the actual json to be served) is cached. More things that make sense to cache should be cached. Care should be taken to only store data under ref/blob id's not refnames/tagnames, but to make refname/tagname requests take advantage of them. Things that come to mind:

• Rendered blobs
• History output
• (Rendered) Blame output
• Patch view

As grabbing data from git is pretty fast, no raw blobs/commits should be cached.