Testing for the behavior of zero length paths, with comments to

Testing for the behavior of zero length paths, with comments to
indicate the preferred behavior.

BUG=398
TEST=Unit testing in PathTest and a new gm test named zerolengthpaths.

[reed1, 2011-12-09 21:29:10]

nice coding structure on the gm

Perhaps we can reduce the matrix somewhat:

1. we can assert (perhaps best as a unittest) that for all combinations of M,
MC, ML, MLC, that all of these never draw filled, and (therefore) always draw
everywhere under inverse-fill. It seems the only test requiring something visual
is stroking

2. we can assert that all stroked paths (these degenerate, and others) always
draw the same regardless of even-odd or winding filltype, and the same
regardless of inverse-even-odd or inverse-winding (since the process of stroking
always generates a winding fillpath).