recOrder buttons working via CLI

[talonchandler]

@ziw-liu @mattersoflight I can demo my progress tomorrow to help us put together a plan.

I've tested this branch against waveorder's pol-refactor branch.

Confirmed working:

• torch installation on hummingbird
• new CLI calculate-transfer-function and apply-inverse-transfer-function can perform all reconstruction steps
• GUI buttons are connected to the new CLI: all 3 acquisition buttons (in both 2D and 3D modes) return results to the screen. Out-of-the-box polarization acquisitions look good. The GUI calls the CLI functions and spits out commands to recreate the transfer functions and reconstructions...initial tests look good.

Major limitations to address:

• phase reconstructions need some tuning...I expect I've introduced a bug on the recOrder side based on my tests, but I'll look closer tomorrow
• background corrections are not connected...I had some issues handling strings vs. paths in the yaml file. @ziw-liu I may grab your help here.
• regularization parameters are not connected and currently use defaults...I had an issue handling a variable number of parameters and passing to kwargs. @ziw-liu if you have capacity I'll grab you here as well.
• I have implemented a very simple file organization scheme (all files go into the snap_directory, and I'm currently recalculating the transfer functions for every acquisition. A better location for these files will enable easy reuse.

Minor limitations:

• haven't implemented an end-to-end reconstruct method yet...this will be a fairly easy last step
• haven't removed much of the unused code in recOrder...commented for now to smooth the transition
• gpu-related buttons are no longer connected
• although I'm saving the transfer functions to file, I don't have easy visualizations working because of minor issues related to real vs. complex values.
• docs, CLI automated testing, much more!
• apply_inverse_transfer_function is where most of the densest recOrder-waveorder logic lives (these recOrder settings require these TFs applied in this way to these waveorder models). I could likely use some improvement to my use of iohub.

[talonchandler]

Merging into 0.4.0dev. Will write TODOs into issues.

[edyoshikun]

Need the following :

from pydantic import BaseModel, validator, DirectoryPath

class _BirefringenceApplyInverseSettings(BaseModel):
background_path: str = ''
    @validator('background_path')
    def check_dir_path(cls, v):
        return DirectoryPath().validate(v)