should test use of implicit derivatives in point rasterization mode

[dneto0]

In point rasterization, it's not so obvious that there will be helper invocations to assist computation of implicit derivatives.

The CTS should exercise this case, particularly the functions mentioned in https://gpuweb.github.io/gpuweb/wgsl/#derivatives

Sending to dj2 as part of the big bag of work to track.

[dneto0]

Kudos to @Kangz for raising this.

[dj2]

@jrprice will this be covered by the derivative tests you've been adding?

[jrprice]

No, I didn't consider different primitives and only tested with a single triangle.