Fix excisor bounds

[nmearl]

Minor fix to bounds in excised region.

Example of bug:

In [23]: from specutils import SpectralRegion

In [24]: from specutils.fitting import fit_generic_continuum

In [25]: spec = Spectrum1D(flux=np.random.sample(100) * u.Jy, spectral_axis=np.arange(100) * u.AA)

In [26]: inc_regs = SpectralRegion(80 * u.AA, 90 * u.AA) + SpectralRegion(50 * u.AA, 60 * u.AA)

In [27]: exc_regs = inc_regs.invert_from_spectrum(spec)

In [28]: cont_mod = fit_generic_continuum(spec, exclude_regions=exc_regs)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-28-a7adc8495e39> in <module>()
----> 1 cont_mod = fit_generic_continuum(spec, exclude_regions=exc_regs)

~/projects/specutils/specutils/fitting/continuum.py in fit_generic_continuum(spectrum, median_window, model, fitter, exclude_regions, weights)
     55     #
     56
---> 57     return fit_continuum(spectrum_smoothed, model, fitter, exclude_regions, weights)
     58
     59

~/projects/specutils/specutils/fitting/continuum.py in fit_continuum(spectrum, model, fitter, exclude_regions, window, weights)
     95     #
     96
---> 97     continuum_spectrum = fit_lines(spectrum, model, fitter, exclude_regions, weights)
     98
     99     return continuum_spectrum

~/projects/specutils/specutils/fitting/fitmodels.py in fit_lines(spectrum, model, fitter, exclude_regions, weights, window, **kwargs)
     64
     65     if exclude_regions is not None:
---> 66         spectrum = excise_regions(spectrum, exclude_regions)
     67
     68     #

~/projects/specutils/specutils/manipulation/utils.py in excise_regions(spectrum, regions, exciser)
    102
    103     for region in regions:
--> 104         spectrum = excise_region(spectrum, region, exciser)
    105
    106     return spectrum

~/projects/specutils/specutils/manipulation/utils.py in excise_region(spectrum, region, exciser)
    150     #
    151
--> 152     return exciser(spectrum, region)

~/projects/specutils/specutils/manipulation/utils.py in linear_exciser(spectrum, region)
     52     flux = spectrum.flux.value
     53     modified_flux = flux
---> 54     modified_flux[s:e] = np.linspace(flux[s], flux[e], len(inclusive_indices)+1)
     55
     56     # Return a new object with the regions excised.

ValueError: could not broadcast input array from shape (51) into shape (0)

In [29]:

[brechmos-stsci]

lgtm

[eteq]

Not a big deal, but minor suggestions for this sort of thing in the future: since the test is specifically for a known bug it's good to reference the issue for future devs. E.g. a comment saying "this is a test for the issue reported in #403), or even call the function test_regression_403. Not worth changing this after the fact, but perhaps for future PRs, @nmearl