Tests for S_min < W

[zenhack]

We should test the scenario where a (natural) split is shorter than the window size; when I wrote the reference implementation this was unspecified, and I think the current implementation disagrees with the current version of the spec.

[zenhack]

Per discussion in hashsplit/hashsplit-spec#23, I think we're going to go back to requiring S_min >= W, so I'm going to close this as obsolete.