Epsie

Notable changes:
* Drops python 2.7 support. Minimum supported version is now 3.6
* 39: (Backward incompatible) Switches to numpy pseudo-random number generators, as `randomgen` has been incorporated into numpy (and is now deprecated).
* Adds documentation and github actions.
* 52: (Backward incompatible) Removes the stability phase from eigenvector proposal, and fixes negative eigenvalues.

Fixes an issue in the eigenvector proposal where eigenvalues could occasionally be negative due to numerical error. Also fixes a few typos.

Adds several new jump proposals, including:

* Eigenvector proposal
* Component-wise scaling for ATAdaptive proposal
* Fast parameters support
* Bounded eigenvector proposal
* Solid angle proposal

Thanks to Richard-Sti for all his hard work adding these.

This release contains many new features. Thanks to Richard-Sti for all the hard work adding these!
* Adds transdimensional, birth/death support.
* Adds `ATAdaptive` proposals. This is an adaptive algorithm that will adapt the full covariance matrix on the fly.
* Adds support for modifying temperatures ("annealing"), with the algorithm from Vousden et al. for adapting temperatures.
* Discrete proposal has ability to return 0 jumps.

Adds Bounded Normal, Angular, and Discrete jump proposals. Also fixes a serious bug in the way the sampler set up the joint proposal, that caused the provided proposals to not be used. That fix slightly changed the way the sampler is initialized (see PR 13 for details). Consequently, this change version is not compatible with previous versions.

Fixes a bug in the way the adaptive normal proposal was being adjusted. See PR 7 for details.