improve: migrate from tensorflow to pytorch

[carynbear]

Tensorflow is used in:

• src/dynadojo/systems/fbsnn_pde/bsb.py
• src/dynadojo/systems/fbsnn_pde/hjb.py
• src/dynadojo/systems/utils/fbsnn.py
• src/dynadojo/baselines/dnn.py

Rather than use multiple ML frameworks, we want to simplify and use either torch or tensorflow. I think torch is the better option because of PyTorch's dynamic computation graph, ease of debugging, and its increasing popularity in the research community.

Once refactored, please test and make sure the demos are working with similar results. And then remove tensorflow packages from pyproject.toml and installation from Docker.

Rebuild and publish Docker images accordingly.

[carynbear]

src/dynadojo/baselines/dnn.py --> refactored to use torch.
tensorflow version is now in src/dynadojo/baselines/dnn_keras.py.

[carynbear]

1. Let's just remove these systems and algorithms from the code and documentation entirely.

• src/dynadojo/systems/fbsnn_pde/bsb.py
• src/dynadojo/systems/fbsnn_pde/hjb.py
• src/dynadojo/systems/utils/fbsnn.py
• src/dynadojo/baselines/dnn_keras.py

2. Remove tensorflow from the pyproject.toml

@carynbear will need to handle docker rebuilding, ignore that for now.