[PATCH 0/7] textconv caching

[Jeff King <peff@peff.net>]

This is replacement for my fast-textconv series from a week ago. It's
much faster, and doesn't require the user to do any additional work when
writing or configuring a textconv helper. The cache is stored as a notes
tree.

On my sample commit, 45 jpegs and avi files with their metadata changed,
and a textconv helper that extracts the metadata, the speedup is:

  [before]
  $ time git show >/dev/null
  real    0m13.724s
  user    0m12.057s
  sys     0m1.624s

  [after (with cache primed)]
  $ time git show >/dev/null
  real    0m0.009s
  user    0m0.004s
  sys     0m0.004s

If you just blinked at those numbers, yes, it really is that much
faster.  The caching dropped it to about .35 seconds, which I showed in
an earlier "how about this" patch. The rest of it comes from patch 7/7,
where we can avoid even opening the binary blobs at all (in my sample,
they total about 180M).

The patches are:

  [1/7]: fix const-correctness of write_sha1_file
  [2/7]: fix textconv leak in emit_rewrite_diff
  [3/7]: make commit_tree a library function
  [4/7]: introduce notes-cache interface
  [5/7]: textconv: refactor calls to run_textconv
  [6/7]: diff: cache textconv output
  [7/7]: diff: avoid useless filespec population

-Peff