AIController helpers, normalization and possibly others

[Ivan-267]

Depending on the env type, a helper might be useful to allow automatic normalization of some features.
For example, linear/angular velocity of an object.

It shouldn't apply to all obs, as normalizing e.g. boolean values converted to float is not needed. Instead, it could be a helper method that perhaps normalizes float arrays (or other types) and returns a value.

We could either add this and potential other helper methods to the AIController that is extended from directly, or make a new class / node for that, which could be referenced from AIController.

E.g. usage:

obs.append_array(helper.get_normalized_array(float_array_to_normalize))

(Actual usage syntax may need to be a bit different to enable capturing min, max, variance, etc.)

SB3 has a wrapper for this, but it applies to training, where a helper function on Godot plugin side would be universal and apply to all frameworks and inference.

I'm just writing this down as an idea that to consider for the future, whether to add this and the specific implementation details and algorithm to use are open to consideration and discussion, the algorithm could be something similar to the sb3 wrapper.

There might be other helpers we can consider, perhaps one to stack obs (although this might be better to implement on the python env and onnx export script as preprocessing, due to less data being transferred), and/or some simple methods for calculating the reward, specifically some things that could be commonly used across envs.