add basic documentation on internal workings of pyjulia

Hello all. This package seems to be rather in need of attention. I've recently made this [post](https://discourse.julialang.org/t/pyjulia-in-desperate-need-of-attention-form-someone-who-knows-what-theyre-doing/2702) on the Julia discourse forum, in which I argue that maintenance of this package seems very important to the future of the Julia language itself. To summarize that argument: there are lots of people who will refuse to ever use Julia if it cannot be called (easily and reliable) from Python. Lots of those same people would much rather use Julia than C or C++ for the performant backend of their Python programs.

Anyway, it's a bit difficult right now for most people with Julia, Python and C experience to see exactly what's going on here. It's pretty clear that this is built around calling `jl_eval_string` from the Julia library, but there are some subtleties (in particular the need for the "fake-julia") that are not so obvious (at least to me). Would someone be willing to write some basic documentation on the internals of pyjulia so it might be a little easier to make contributions? Thanks.