Migrate fluorescence reconstructions to `torch` and `model`-based structure about waveorder

[mattersoflight]

Yes, isotropic_fluorescent_thick_3d is immediately useful - please prioritize that while keeping in mind the overall dependencies between different models.

[mattersoflight]

@talonchandler Here is my reading of inputs and outputs, and suggestions for naming conventions:

2D -> isotropic_fluorescent_thin_3d: 3D input is a short z-stack (say 5 slices covering the DOF) and output is the restored fluorescence at the plane of focus using defocused OTFs. Analogous to isotropic_phase_thin_3D.
3D -> isotropic_fluorescent_thick_3d: 3D deconvolution.

We should also add isotropic_fluorescent_thin_2D, which would be a simple 2D deconvolution with 2D OTF.

The most basic (geometric) model for fluorescence anisotropy Fluor_anisotropy_recon may be named inplane_orientation_fluorescent_thick_pol3d. The input is a 3D polarization-resolved stack and the output are 3D as well. I am suggesting orientation over the word fluorescence anisotropy because anisotropy signifies rotational diffusion to many.

[talonchandler]

Great, I like these suggestions a lot.

isotropic_fluorescent_thick_3d is the highest priority, and the following three can follow later:

isotropic_fluorescent_thin_3d
isotropic_fluorescent_thin_2d
inplane_oriented_fluorescent_thick_pol3d

[talonchandler]

The highest-priority fluorescence reconstruction isotropic_fluorescent_thick_3d was migrated to torch in #128.

I'm collecting the remaining migrations into a list here:

• isotropic_fluorescent_thin_3d
• isotropic_fluorescent_thin_2d
• inplane_oriented_fluorescent_thick_pol3d