ComputeGeometryX rework

[jaharris87]

This PR contains 2 things:

• Add optional arguments to ComputeGeometryX to allow computation on a local mesh and geometry, independent of the global MeshX and CoordinateSystem from modules.
• GPU-ize ComputeGeometryX so that it can be called without G already being on the GPU, as would have been the case with the existing update directives. A side effect of this is that G will not be up-to-date on the host unless copied over elsewhere---I think this is okay since I don't believe we use G anywhere on the host except in diagnostics or I/O, where it is already being explicitly updated.

@dunhamsj I think this PR likely effects you the most, so could you review my changes an maybe test this to make sure you're still getting the same/correct answers?

[jaharris87]

Also, I'd accept feedback from @dunhamsj , @eendeve , or others that see this if there might be a better way to do this. Particularly, I made a choice to pass in MeshX, but maybe just passing element centers and widths would be better?