Pyvibdmc

This release makes it so that the PyVibDMC API is much cleaner than the original implementation. I now recommend to use `import pyvibdmc as pv` for all interactions with the package.

Additionally, the introduction of CuPy descriptors is formalized in the code, but still needs to be documented. This will be added in a later release. The descriptors are CuPy/NumPy interoperable, meaning CuPy is not required to use them. These include the (sorted and unsorted) distance matrix, the coulomb matrix, and the SPF matrix (delta R / R).

Bug fixing in the molecular rotation code, the xyz to npy code, and the bisector function in the analysis code. Prepped for DOI allocation through Zenodo

For NN DMC, it is important to have fast transformation from cartesian to the molecular descriptor of interest, in this case the coulomb matrix. As such, we made an additional implementation using CuPy, which enables the use of NumPy syntax to get GPU-accelerated code / CUDA efficiency. Eventually, CuPy will be a requirement for running NN DMC, but for now tensorflow is still the only requirement.

No significant impact on user-facing code, but CI has been successfully migrated to use GitHub actions as opposed to Travis-CI. Codecov still works as well.

In this release, there is now a flexible `NN_Potential` object that one can use instead of a `Potential` object in the `simulation_utilities` that allows you to run your code with a tensorflow keras neural network model. Please refer to the documentation.

Additionally, the continuous integration framework will soon migrate to GitHub actions instead of Travis CI, as it has unlimited free use unlike Travis.

Install the latest release of PyVibDMC using `pip install pyvibdmc`

In this release, we added a `InitialConditioner` class that handles the initialization of the DMC ensemble before starting the DMC simulation. Typically, one can just pass in the minimum energy geometry blown up by a certain factor (geom * 1.01). However, in some cases, it is advantageous to start with a pre-sampled ground state, or at least initialize your walkers to permute like atoms. In this new release, we have given users the capability to do both of these things. Please see the new section in the [Documentation](https://pyvibdmc.readthedocs.io/en/latest/).