Add VGC teampreview functionality into environments

[cameronangliss]

I've been sitting on this code for a while but I didn't want to contribute it until I could figure out a minimal implementation, but now I have so here you go! This adds VGC teampreview as part of the MDP, overriding the previous default random action that was taken at teampreview for agents in the environment. The total teampreview, which looks like "/team wxyz" where w, x, y, and z are numbers from 1 to 6, is determined by two doubles orders given by the agent. The first order gives the Pokemon to choose in teampreview that are represented by w and x, and the second order gives y and z.

[codecov]

Codecov Report

❌ Patch coverage is 21.62162% with 29 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.13%. Comparing base (f458350) to head (356ebf6).
⚠️ Report is 239 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #780      +/-   ##
==========================================
- Coverage   83.38%   83.13%   -0.25%     
==========================================
  Files          39       47       +8     
  Lines        3918     5628    +1710     
==========================================
+ Hits         3267     4679    +1412     
- Misses        651      949     +298     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[cameronangliss]

@hsahovic this is ready for review whenever you get a chance!

[hsahovic]

I could be convinced this makes sense, but my understanding is that this is trying to rely on choose_move for teampreview, which I'm not convinced is correct. Why do you think we should generally do it that way?

[cameronangliss]

@hsahovic I'm doing that because choose_move is gathering decisions from the _EnvPlayers, which is the action input from env.step(). Basically this allows the RL agent to control the battle during teampreview using switch actions. One could do it differently I suppose but I see no reason why not to do it like this, since the switch action translates 1-to-1 to the teampreview decision case (you just need to make two double-switch decisions in a row to get a teampreview decision of length 4).