Draft: some improvements on separate init branch

[jat255]

Per our conversation on Friday...

• Slicing with a single location with isig now emits a warning, and coerces the shape to (ntilts | 1, y) :

>>> s = ds.get_catalyst_data(misalign=False)
>>> s.isig[20, :]
Slicing a TomoStack signal axis with a single pixel is not supported. Returning a single pixel on the "x" axis instead
<TomoStack, title: , dimensions: (90|1, 600)>

>>> s.isig[:, 200]
Slicing a TomoStack signal axis with a single pixel is not supported. Returning a single pixel on the "y" axis instead
<TomoStack, title: , dimensions: (90|600, 1)>

• Added an extract_sinogram() method that supports both integer and float-based indexing:

• Fixed some axes-handling issues that were causing tests to fail.

Todos:

• Test/check test output of CUDA code
• Add coverage for RecStack constructor
• A few other little things I found

But wanted you to see this so we don't duplicate efforts

[jat255]

Closing in favor of #18, which is direct to master