Tkurth/interface fixes

[azrael417]

This MR adds some more tests for shapes of tensors emitted from replay and rollout buffers and also multi env tests for replay buffer.

This MR fixes shape checks in the training routines for various RL algorithms

This MR adds Debug compilation mode to Dockerfile

[romerojosh]

I pushed a commit to fix up some of the Fortran to C boolean conversions in the interface as we discussed.

I also noticed that there was some inconsistency in the names of arguments, particularly final_state and terminal across the Fortran and C APIs. I went ahead and just converted them all to final_state but feel free to change them all to terminal if you think that is better. The important thing is that we are consistent everywhere.