Try maintaining blame information internally

[asharov]

Most of the time of commit processing is spent on running git blame on the changed files. I've tried a few things to speed this up, but it seems like there are no easy major improvements to be had here.

One possibility would be to avoid git blame completely, except for the first commit. Since Git Hammer is processing the commits in order, it could track, in addition to the statistics it collects, also complete per-line author information for all the files. This would make it possible to determine a new commit's statistics based only on the diff information without having to run blame.

It's probably too much to save the complete line-by-line author information in the database. This would be more an optimization during processing a large number of commits. But it could be beneficial to save this information for the head commit, so that an update could start processing in the same faster way as the existing commits. This should be experimented with, but it's not mandatory to resolve this issue.