Use .value to avoid error calling nan_to_num on quantity

[keflavich]

Using a cube with BUNIT = 'K', I ran into the following traceback:

  File "/Users/adam/repos/pvextractor/pvextractor/pvextractor.py", line 97, in extract_pv_slice
    order=order, respect_nan=respect_nan)
  File "/Users/adam/repos/pvextractor/pvextractor/geometry/slices.py", line 42, in extract_slice
    slice = extract_line_slice(cube, x, y, order=order)
  File "/Users/adam/repos/pvextractor/pvextractor/geometry/line_slices.py", line 59, in extract_line_slice
    total_slice = map_coordinates(np.nan_to_num(cube), [zi,yi,xi], order=order, cval=np.nan)
  File "/Users/adam/anaconda/envs/astropy27/lib/python2.7/site-packages/numpy/lib/type_check.py", line 379, in nan_to_num
    y[are_inf] = maxf
  File "/Users/adam/repos/astropy/astropy/units/quantity.py", line 807, in __setitem__
    self.view(np.ndarray).__setitem__(i, self._to_own_unit(value))
  File "/Users/adam/repos/astropy/astropy/units/quantity.py", line 1023, in _to_own_unit
    raise exc
UnitsError: '' (dimensionless) and 'K' (temperature) are not convertible

pvextractor works on ndarrays; it has no reason to concern itself with flux units.

[astrofrog]

Looks good! Can you add a regression test?

[keflavich]

We actually never had tests using SpectralCube objects, so those are needed.

[keflavich]

@astrofrog - check it out, class-level parameterization.

[keflavich]

These test failures probably result from my having a more recent SpectralCube version that parses units appropriately. Oops.

[keflavich]

@astrofrog - I think this is ready to merge; go ahead if you agree

[astrofrog]

Looks good, thanks!