Create Boltzmann velocity distributions from `Simulation`s directly

[Alexsp32]

Fixes https://github.com/NQCD/NQCDistributions.jl/issues/9 and requires https://github.com/NQCD/NQCDistributions.jl/pull/12 and https://github.com/NQCD/NQCModels.jl/pull/35 to be merged prior.

Since the VelocityBoltzmann is dependent on the atomic masses, it makes sense to be able to pull both these and the overall dimensions from an existing Simulation. For consistency, a temperature must still be provided separately, but this is quick to do with Simulation.temperature(0).

Also, initialising VelocityBoltzmann from a simulation should check for any atoms not included in mobileatoms, as these should be initialised with zero velocity.