Rethinking wavelength sampling and wavelengths in the synthesis API

[ajwheeler]

There are two things that could be improved about the way wavelengths are currently handled.

1. Log-uniform wavelengths (uniform velocity steps) provide a better sampling for broad wavelength ranges. It makes for much saner calculations from the UV to the IR, and makes, e.g. Synthetic photometry #261 easier/faster. It also makes convolutions with velocity kernels really mathematically nice, and potentially faster using FFTs.
2. Having to pass a vector of Julia ranges in order to get non-contiguous wavelength ranges is a big pain if you are using python.

My current thinking is that this is worth doing, but I'm going to let the idea cure for a bit.

[ajwheeler]

This will be closed by #336, if it's merged.

[ajwheeler]

I'm leaving this open for now. #336 does most of this, but doesn't make all the breaking changes that are required for me to move to log-uniform wavelength sampling without anyone noticing.