Inefficiency of applying the correction during saving a new h5

[lambjames18]

As of now, the algorithm creates a new set of solutions for each EBSD data modality. This is incredibly inefficient, and needs to be changed.

• core.py needs to actually support a 3D application instead of just a hack
• class core.Alignment needs to be more efficient when initialized. It does not need control points immediately, just the algorithm to be used.

I'm sure theres more, but the jist here is that core.py is inefficent and that makes the UI inefficient. Things are tangled and need to be made modular.

[lambjames18]

Steps to fix:

• Buffer the control data if any control dimensions are smaller than the distorted dimensions (output must have the same dimensions as the distorted data).
• Create a dummy volume of True's
• Align the dummy volume and save the solutions that are generated for the full stack
• Determine the volume crop that will match the dimensions of the output with the distorted data by locating the centroid of the corrected dummy mask and creating cropping slices to decrease the size.
• Loop over the actual data, align it, crop it, and store it.
• Save the aligned and cropped data back to the H5 file

Note: the reason we need to enforce output.shape == distorted.shape is because Dream3D doesn't allow us to change the shape of the data containers without creating a while new data file (and even that is not supported in python currently).