Add MaskingNodes for different types of tiles

[jisantuc]

Overview

This PR turns MaskingNode into an ADT with options for different arities of source data (unary, ternary, quaternary) and adds a failing test where my guess about what should have happened as a result turned out to be wrong.

The reason for this is that we currently can't mask any kind of non-single band imagery result, which is a drag, because in RF for single band viz of mosaics we return a multiband tile of the RGBA channels after rendering, and for multiband viz of mosaics we return an RGB tile.

Testing

• I added a test to ResultSpec that shows that evaluation of MaskingNodes, even after accounting for the possibility of several source bands, squashes everything down to a single band

[jisantuc]

The test I added doesn't currently pass, and I'm not sure why -- evaluation of Masking directives squashes everything down to a single band, but I didn't figure out where / how that's happening.

[jisantuc]

Superseded by #89