spherical volume rendering refactored

[chrishavlin]

So this is a verion of #64 that seems to work pretty well with a few caveats.

This version does 2 things differently:

1. It calculates the cartesian bounding boxes of every block initially (using code from ENH: cartesian cutting planes through non-cartesian geometries  yt#4847), and passes that info down to the shaders as additional vertex attributes.
2. And then those cartesian bounding boxes are used in the ray tracing shader for calculating ray intersections. At each sample point along the ray, the spherical coordinates are calculated and used to sample the texture if it falls in the actual spherical element.

The main caveat is that you need a fairly large number of sample points for the ray tracing since intersection with the cartesian bounding box does not guarantee intersection with the enclosed spherical element.

EDIT: cartsian bbox calculation now works with big elements (via a recursive division when its needed). Also, at least one thing that should be fixed before this goes in is that the cartesian bounding box calculation was designed for fairly small spherical elements and breaks down when blocks span whole quadrants of the spherical coordinate system. e.g., the new example script uses load_uniform_grid with nproc=128, if nproc is smaller (less than 8 or so), the blocks get too big and the bounding box calculation breaks down and results start to get wonky (or nonexistant). So, should fix that...

Finally -- I decided to go with a new PR rather than modifying #64 because I'm abandoning the complicated intersection checks with the underlying spherical element bounds. While I think that approach could work (and could be added on here...), I think it's worth getting the simpler version here merged first and wanted to preserve #64 in its current state to make that attempt easier in the future.

[chrishavlin]

merged! can now move all the geometry-specific calculations to pre-processor directives rather than uniforms.

[chrishavlin]

pre-processor directives now working

[chrishavlin]

still do do:

• docs docs docs
• another test or two would be nice -- maybe a series that checks different bounds.
• an example with a real dataset
• refactor the new cython a bit

[chrishavlin]

note: this is now in conftest.py

[chrishavlin]

copied from test_yt_idv.py so it can be used in other test files

[chrishavlin]

Latest push includes some new logic to limit the options that show up in the GUI so that unsupported shaders and scene components are not shown. I also bumped the maximum sample factor when the data is in spherical coordinates.

[chrishavlin]

Working on docs now...