Use procgen

[dfilan]

Description

Import procgen and add configs to train_rl and eval_policy for procgen environments.

This is basically a very pared-down version of #563 that just enables things that are immediately useful to me.

Testing

Added a test that train_rl works on coinrun, the test passes.

[codecov]

Codecov Report

Merging #639 (1ee83a5) into master (1dd4c8f) will decrease coverage by 0.33%.
The diff coverage is 70.47%.

❗ Current head 1ee83a5 differs from pull request most recent head 245e881. Consider uploading reports for the commit 245e881 to get more accurate results

@@            Coverage Diff             @@
##           master     #639      +/-   ##
==========================================
- Coverage   97.51%   97.17%   -0.34%     
==========================================
  Files          85       85              
  Lines        8316     8421     +105     
==========================================
+ Hits         8109     8183      +74     
- Misses        207      238      +31     

Impacted Files	Coverage Δ
src/imitation/scripts/config/eval_policy.py	72.72% <66.66%> (-3.99%)	⬇️
src/imitation/scripts/config/train_rl.py	75.00% <68.75%> (-3.95%)	⬇️
tests/scripts/test_scripts.py	100.00% <100.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[dfilan]

Procgen has some issues with reward learning, so it probably doesn't make sense to merge this into imitation. (Also it's kind of interesting that tests seem to be 'failing' due to timing out as a result of adding new tests?)