Investigate removing nullmask from parse_uri results when not needed

[hyperbolic2346]

As brought up in review, the nullmask is potentially unnecessary. If the input doesn't have a nullmask the result could still need one. This doesn't mean that it will though as the input URI's could all be valid results.

This investigation is to weigh the performance impact vs the memory savings of removing the nullmask when it is possible. This could require marching the mask afterward or having an atomic the kernel updates based on nulls.

Originally posted by @nvdbaranec in #1704 (comment)