Refactor diffusion solvers

[zhardy-lanl]

• [ ] Move all diffusion related implementation into the diffusion module.
• [ ] Make all diffusion solvers multi-group
• [ ] Modify diffusion-based acceleration schemes to create diffusion solvers from the diffusion module.

Refs #41

[andrsd]

From pr #328:

Wish list:

• the code can solve the diffusion approximation, with CG and DG FEM. This could be useful to initialize a transport (k- or source-driven) solve but it could also be the only solve a user does if they are ok with diffusion. Whether this leaves inside the lbs module or not is a design decision that I likely have no preference for.
• the ability of doing a manufactured solution is very important to me. So having a diffusion solver with the capability is important to me. I found bugs in the MIP-DSA when I coded the standalone DFEM Diffusion solver. There could be a few left. I really need to stress test the DSA and compare the results from my previously published results. I have found the OpenSn DSA to be less effective than what I used to experience in my standalone codes.
• The multigroup diffusion code already has the two-grid acceleration. This is useful and should not be lost.
• I contributed a lot to this mess and I apologize. If the design decision is to keep a diffusion solver outside of lbs and diffusion accelerators inside of lbs, then doing manufactured solutions for the accelerators is not required. Some code may be duplicated but I am not sure if this is tolerable or not.