[Piglit] [PATCH v4] tests: add nvidia conservative rasterization tests

[Pending Chaos]

No, I do not.

I encountered a similar issue where the normal-line test renders differently
with and without -fbo when using nouveau. Looking at the differences, it
seemed the vertices were moved by a small, probably subpixel, amount.
I ended up just skipping the normal-line test when using -fbo (it was not
that important of a test anyways).

The nouveau driver and the nvidia driver also seem to render the triangle
slightly differently when the horizontal subpixel precision bias differs from
the vertical one. This was also worked around.

What exact subtest(s) fail?