Add support for multiprecision arithmetic so we can use very high order maps

[fbartolic]

Although this is definitely not a priority, it would be cool to be able to use Starry very high order maps (l=50 or so) in emitted light for the purpose of inferring high resolution emission maps of Io. Since we know the geometry of these occultations very accurately, we can precompute the design matrices so the implementation wouldn't need to be particularly efficient.

The science goal here is to be able to infer sizes and shapes on hotspots on Io.

[rodluger]

Cool. This was implemented in version 0.3.0 but I haven't been able to get that to compile recently (must be conflicts with newer versions of gcc or something). I'll try to port that functionality over when I have some free time (timescale a few weeks, probably).