Remove CBrock, CBrockDisk, Hernquist

[michael-petersen]

This branch removes the bases that a superseded by more adaptive versions. For the Clutton-Brock (1973) and Hernquist & Ostriker (1992) basis, this is sphereSL; for the Clutton-Brock (1972) disk basis, this is flatdisk.

[The9Cat]

This looks fine to me. And I've checked that it compiles without incident. If you have nothing else to add here, let's merge it.

[michael-petersen]

I'm thinking we want to keep bessel, as a unique basis? That was my only open question.

[The9Cat]

It might be worth keeping bessel because it provides something that the others do not. Athough it does not have pyEXP integration or cuda support. Neither of those would be all that hard to do. We might want to try it to make sure it works correctly...

[michael-petersen]

Right, I agree if we keep bessel (probably the right call) we should have some sort of test. Any thoughts on what this would be?

pyEXP integration is a bonus, but could also be a minor rev later if requested.

[The9Cat]

For a test, perhaps:

• Realize a King model and make sure it's stays in equilibrium
• Use Oskipkov to make a radial orbit instability

Yes, adding to pyEXP would be an hour of work. Not a big deal. CUDA would be similar. Part of the work is already done because the bessel implementation uses tables, and that's the same as Sphere, so it can be leveraged. Thinking about that, the right thing to do is to move the CUDA Sphere implementation to SphericalBasis and use the same code for both.

[michael-petersen]

As the bessel fixes are being implemented in #73, I'll merge this in now.