Add support for dimensionless spectral regions

[duytnguyendtn]

Dimensionless unit SpectralRegions would seem to suffer from the same bug as #846. In conversations with @rosteen, there is a fair debate about whether SpectralRegions should support this usecase (considering... they're not spectral units). This issue is meant to encapsulate that discussion. Other interested parties I'm pinging: @eteq @pllim @camipacifici @orifox @PatrickOgle

[pllim]

We need a data container for uncalibrated spectrum, right? Seems like a good package to have this support, which specreduce can also use.

[nmearl]

I suppose the big question is whether a spectral region is still a relevant physical description of an area of a spectrum without knowing the dispersion unit. My first reaction is that it doesn't make sense to define a spectral region on a spectrum with no units; I'd assume you'd still need something to describe the dispersion.