Add vectorized gym support and other improvements

[engintoklu]

New features:

• VecGymNE: a new problem type for vectorized gym environments. VecGymNE supports brax environments and can do its computations on a GPU.
• Fixed and simplified recurrent network policies for both GymNE and VecGymNE. From now on, torch modules which accept a second optional argument and return a tuple of two elements (output, hidden_state) will be treated as recurrent policies. Modules with reset(...) method will NOT be treated as recurrent anymore.
• str_to_net(...): Deprecated the names RecurrentNet and LSTMNet. One is encouraged to use the names RNN and LSTM instead.
• New environment variable EVOTORCH_TOP_IMPORTS. For example, when EVOTORCH_TOP_IMPORTS is set as "pickle,ast", the modules pickle and ast will be imported in this order at very early stages, both by the main process and at remote actors. This feature can be used when certain libraries need to be imported before others.
• PicklingLogger: a new logger which periodically saves the current solution. When working with GymNE and VecGymNE, saves the current policy (as a torch module) and the observation stats (if any) as well.
• The visualize(...) method of GymNE is updated so that it works with the new rendering API of gym 0.26.
• The example script rl_enjoy.py is updated so that it works with the new rendering API of gym 0.26.
• Added an example notebook demonstrating how one can solve the brax humanoid task using VecGymNE.

[codecov-commenter]

Codecov Report

Merging #20 (6f83f31) into master (df034a4) will decrease coverage by 1.17%. The diff coverage is 44.30%.

@@            Coverage Diff             @@
##           master      #20      +/-   ##
==========================================
- Coverage   54.66%   53.48%   -1.18%     
==========================================
  Files          35       43       +8     
  Lines        4978     6254    +1276     
==========================================
+ Hits         2721     3345     +624     
- Misses       2257     2909     +652     

Impacted Files	Coverage Δ
src/evotorch/neuroevolution/net/rl.py	20.49% <20.77%> (-4.51%)	:arrow_down:
src/evotorch/neuroevolution/vecgymne.py	20.89% <20.89%> (ø)
src/evotorch/logging.py	28.82% <22.22%> (-7.93%)	:arrow_down:
src/evotorch/neuroevolution/net/misc.py	64.70% <22.22%> (+30.09%)	:arrow_up:
src/evotorch/core.py	66.53% <23.07%> (-0.47%)	:arrow_down:
src/evotorch/neuroevolution/gymne.py	26.69% <24.70%> (-0.81%)	:arrow_down:
src/evotorch/neuroevolution/net/statefulmodule.py	36.36% <36.36%> (ø)
src/evotorch/neuroevolution/net/vecrl.py	40.06% <40.06%> (ø)
src/evotorch/algorithms/searchalgorithm.py	38.82% <45.45%> (+0.13%)	:arrow_up:
src/evotorch/neuroevolution/neproblem.py	35.16% <53.84%> (+1.42%)	:arrow_up:
... and 17 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[flukeskywalker]

Let's merge #25 to this branch, then merge this.