Switch ObjectWalk to use a naked CanonicalTreeParser because its faster
Removing the indirection of the TreeWalk during an ObjectWalk's loop
over the trees saves us 20 seconds on one of my linux-2.6 clones. On
average we went from 1m37s for "jgit rev-list --objects" down to 1m17s.
C git still does the same operation in 0m20s, so we're a long way from
matching C git's performance, but this is a worthwhile improvement, even
though the code has become slightly more complex.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Robin Rosenberg <robin.rosenberg@dewire.com>