Before this commit foo_1, foo_2, ..., foo_10, foo_11 would be sorted
as foo_1, foo_10, foo_11, foo_2, ... (i.e. lexicographically).

After this commit, they are sorted naturally (per human intuition),
as foo_1, foo_2, ..., foo_10, foo_11.

As such, there should be no more need to name files as foo_0001,
foo_0002, etc, so long as the files use numbers of ascending
sequence it should work just fine.

This issue was reported by @nickgavran.