Add HMCMove

[maxentile]

• Add HMCMove
• Make misc. small improvements to moves.py
• Test HMCMove
  • Test stability etc. on vacuum ligand (using Jax potentials)
  • Test stability etc. on solvated DHFR (using timemachine potentials)
  • Test accurate sampling on system of independent oscillators (where 1D marginals can be computed ~exactly and compared to histograms)

---

Note: Standard HMC is not a practical move to apply directly to large MM systems, but it may be practical to apply to small systems (or subsystems).

(To get an acceptance rate of > 50% on the DHFR test system, for example, we need to reduce the timestep to an eye-watering 0.01 fs or so. A variety of mitigating strategies have been proposed in the literature, which may be necessary to consider if we ever need to use "global" HMC moves on these systems.)