Merge develop branch into master (2022/1)

[ikki407]

This update is quite large. We must validate a performance with some experimentation before merging to master branch.

All Changes

• feature: link the 1st place solution in Hungry Geese competition #235
• feature: explain DRC net #238
• feature: (idea) create observation mask from observation!=None #240
• fix: there is cases if we have no value outputs nor policy outputs #242
• feature: true drc #237
• feature: use load_model() function #231
• chore 2022/01 #236
• fix: trainer for non trainable model #239
• (Idea) feature: init hidden torch #243
• feature: add make_onnx_model.py #251
• feature: add scripts/aux_swa.py #252
• feature: hidden hidden argument count #253 (maybe Revert "feature: hidden hidden argument count" #257)
• chore: update description of observers() #250
• feature: assign cumulative worker index #246
• chore 2021/01 2nd #256
• (WIP) feature: ONNX support for evaluation #233
• (To be discussed) feature: store selected probability only (IMPORTANT) #232
• (To Be discussed) (RNN) feature: burn in steps #223
• (Idea) feature: unnecessary hidden argument #218
• (WIP) feature: add observers() method in environments #216
• (Idea) feature: no observation #202
• feature & fix: batch_shape, flatten, unflatten, P or 1 #260
• feature: add scripts/win_rate_plot.py #258

To be considered in the next update

• (Idea) feature: update configuration for turn-based #241
• feature: cumulative worker id without entry server #255
• (Idea) feature: set optimizer from outside #193
• multi-dimensional reward
• availability of entry-server