ML state estimation doc

[matthewware]

ML state estimation seems to only work with 6 observables. In the case of four observables {+Z, +X, +Y, -Z} the reconstruction seems to struggle. This might intuitively make sense when reconstructing a density matrix of 4 or more parameters. Just from anecdotal experience, I'd like to update the README and doc to recommend using more observables for the ML reconstruction. This seems to give better quality reconstructions.

[matthewware]

So four observable don't form a POVM which is required by the ML reconstructions. This is true at least for the four observables we (BBN) use in QGL. The right place to put this doc is probably in Qlab.jl where the problem gets setup.

[matthewware]

See the code and examples in Qlab.jl and documentation. The use of the ML reconstruction code for actual data, especially in the two-qubit case is delicate and depends a lot on the data.