Implement optional Jax backend

[mtessmer]

Currently, chiLife uses numba as a cross-platform jit compiler to improve performance for computationally expensive tasks. Unfortunately, these numba jitted functions can not be easily used with any auto differentiation packages, which would enhance several modeling procedures, particularly bifunctional rotamer modeling, and minimization. Adding an optional Jax backend would allow for auto differentiation of many computationally demanding procedures, enhancing the performance of gradient-dependent functions. Computational efficiency would be maintained through XLA compilation and Jax would even allow for improved computational performance by utilization of GPU acceleration when available.