More thorough/advanced line-flux formulation

[eteq]

This issue builds off of discussions in #235, #314, and #324.

To summarize: #324 implemented a "quick and dirty" way of doing line flux measurements - basically it only considers whole-pixels. The "right" way is to also include the fractions of pixels that are on the edges. For a more in-depth discussion of this with math and everything see https://github.com/astropy/specutils/issues/314#issuecomment-421109230 .

This issue is to improve the line_flux function to account for the fractional edge pixel. I think the linked comment above lays that out reasonably thoroughly, but if more clarification is needed/changes proposed, we can hash it out in this issue.

By reference this will also affect equivalent_width, SNR, etc., I think which is a good thing.

[nmearl]

@eteq can you expand on how this might be implemented? If we have a spectrum like

spec = Spectrum1D(flux=[10, 5, 3, 7] * u.Jy,
                  spectral_aixs=[1, 2, 3, 4] * u.um)

and want the flux between 2.5 u.um and 3.25 um, would we interpolate onto a finer grid before doing the summation?