refactored multi agent proposal

[Mytolo]

Instead of returning a generator for RLBase.plan! use NamedTuples s.t. the environment is able to gather the actions for the players correctly. It might be that the iteration of the environment is not in order (for example, this might occur when it is wrapped from another language into julia).

Moreover, this PR includes proper push operations for the Basic Agent policy for simultaneous dynamicStyle trait envs. They were missing. Hence, the trajectory was not updated acordingly.

[codecov]

Codecov Report

Merging #926 (ec34889) into main (1f7f347) will decrease coverage by 13.65%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##             main    #926       +/-   ##
==========================================
- Coverage   13.64%   0.00%   -13.65%     
==========================================
  Files         217     208        -9     
  Lines        7571    7415      -156     
==========================================
- Hits         1033       0     -1033     
- Misses       6538    7415      +877     

Impacted Files	Coverage Δ
...inforcementLearningCore/src/policies/agent/base.jl	0.00% <0.00%> (ø)
...mentLearningCore/src/policies/agent/multi_agent.jl	0.00% <0.00%> (ø)

... and 42 files with indirect coverage changes

[Mytolo]

NOTE: CI fail depends on rework of run loop as RLBase.plan! is executed after termination of environment as stated in #913, PR #921

[HenriDeh]

Doesn't the new loop work for you ?