Improve interface of BasePolicy.compute_action

[opcode81]

• I have marked all applicable categories:
  • exception-raising bug
  • RL algorithm bug
  • documentation request (i.e. "X is missing from the documentation.")
  • new feature request
  • design request (i.e. "X should be changed to Y.")
• I have visited the source website
• I have searched through the issue tracker for duplicates
• I have mentioned version numbers, operating system and environment, where applicable:

  import tianshou, gymnasium as gym, torch, numpy, sys
  print(tianshou.__version__, gym.__version__, torch.__version__, numpy.__version__, sys.version, sys.platform)

The interface is not optimal, because not all observations are indeed arrays. Observations can also be array-like only (numpy's ArrayLike type). In fact, this is the case for some of gymnasium's standard implementations (e.g. LazyFrames).

I am not sure how appropriate the current type annotation is (i.e. arr_type, which also includes Tensor).

Proposal:

• Change type annotation of obs to numpy's ArrayLike
• Add first line which ensures array representation: obs = np.array(obs). This will convert anything array-like to an actual array (e.g. using an implementation of __array__ if necessary).

[MischaPanch]

Closed by #1170