Add capability of threading over IntRange ranges, use in l2_norm_diff()

[roystgnr]

This request turned out to be harder than it should have been, but if it works then there's a ton of other opportunities that ought to be much easier to make threaded now that IntRange has been fixed up to do that more easily.

[moosebuild]

Job Coverage, step Generate coverage on a0b7938 wanted to post the following:

Coverage

	325622	#4446 a0b793
	Total	Total	+/-	New
Rate	65.47%	65.48%	+0.01%	100.00%
Hits	78228	78256	+28	28
Misses	41258	41253	-5	0

Diff coverage report

Full coverage report

This comment will be updated on new commits.

[roystgnr]

@NamjaeChoi tested this and reports semidecent speedup (37.6s in serial, 8.0s on 8 threads). This is in NumericVector so the awful virtual operator() overhead dominates everything and a real optimization would probably be a PetscVector specialization, but at least we're confirming that it works and there aren't any massive weird thread contention issues, even with the tricky PetscVector::_get_array locking.