stash: avoid unnecessary reset_tree() call

In 45c1389 (stash: convert apply to builtin, 2018-12-20), we
introduced code that is the equivalent of `git write-tree && git
read-tree`. But the original shell script only called `git write-tree`
(because the read-tree would obviously be a no-op). So let's skip the
reset_tree() call that is the equivalent that that `git read-tree`.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>

builtin/stash.c

@@ -404,7 +404,7 @@ static int do_apply_stash(const char *prefix, struct stash_info *info,
 	if (refresh_cache(REFRESH_QUIET))
 		return -1;
 
-	if (write_cache_as_tree(&c_tree, 0, NULL) || reset_tree(&c_tree, 0, 0))
+	if (write_cache_as_tree(&c_tree, 0, NULL))
 		return error(_("cannot apply a stash in the middle of a merge"));
 
 	if (index) {