Consider doing hunk-level diffs in the ObserverEngine

[Notgnoshi]

The diffs that the current RuleEngine does are file-level, but you still have to take a file and perform the diff yourself to get the actual hunks. Right now, that's deferred to the Rules, but I expect that many observers will be interested in the hunks, and not just the file Changes.

[Notgnoshi]

I don't have any Observers that actually need to diff the changed files, so I need to add one first before I can figure this out.

[Notgnoshi]

See #146 - it looks like the blob fetching is the most expensive part of the current implementation, so if that can be pre-fetched from the ObserverEngine thread, then we can

• utilize the pack-cache-lru-dynamic cache (in a linear commit iterator this cheapens every commit diffed?)
• prevent each observer from having to pay this cost (there's only one that pays it right now)