Implementation of extractions at different wavelengths per fiber

[segasai]

This is an initial implementation.
First get_spots() is updated to support 2d wavelength arrays. The tests were added to check we get the same results with 2d array as with 1d array.
The next step is to figure out what's the best way to update the extract_frame() interface.

[segasai]

After creating a test implementation of the extract_frame() that supports 2d wavelength array, i realized that since extract_frame returns a desispec.Frame (which only has a 1d wavelength array), so it wouldn't quite make sense to just allow 2d wavelength input for extract_frame.

The two alternatives for extract_frame() is

1. Allow 1d wavelength_array input, and optional velocity offset array per fiber
2. Allow 2d wavelength array, and additional 1d wavelength array (that will be in the in the returned desispec.Frame), i.e. this is for the case where 2d wavelength array is the telescope frame wavelengths, and the 1d frame is the barycentric frame grid on which the result will be.

I think 2 is cleaner.