Ezff

Major release with new features and bug-fixes

Features:
1. Includes several new gradient-free optimizers from multiple frameworks -- Nevergrad, Pymoo, Platypus, and MOBOpt. Updated documentation to reflect new algorithms
2. Better LAMMPS support, both for spawning LAMMPS jobs and reading LAMMPS dump/output files
3. Better error handling -- Errors now generate `np.nan` objectives, instead of crashing the optimization routine
4. Ability to save and restart from previously evaluated variable sets
5. More examples for serial/parallel forcefield parameterization using LAMMPS/GULP, and parameterization of hybrid forcefields
6. Parallelization using both multiprocessing and mpi pools

Preparing for v1.0 -- to be released after minor documentation updates

Release after a minor rewrite and refactoring

New features:
1. OOP-ed error calculation and parsing of MD runs and ground truths
2. Parameterization of ReaxFF charge terms
3. New error measure for structural distortion
4. Forcefield handling routines are less cumbersome to use
5. New examples and updated doc page

Bugfix: Resolves missing variable errors during NSGA-III initialization

Feature: Easy way to use between multiple algorithms during a single optimization run