Enhance README

[Anakin100100]

I recently reanimated my laptop with an Nvidia GPU and installed Ubuntu there. I noticed this project and wanted to play with it but the documentation is quite sparse on the setup which can deter potential contributors and users.

This PR enhances the README with a step by step instructions on which software needs to be installed and how to use PufferLib to train and export a neural net. I tested the setup on a fresh container. The expected observations and comments on the training process are based on my limited knowledge of RL so I'd welcome your guidance there.

The setup instructions are centered on Ubuntu but provide alternative instructions for all major pieces of software that are required for other Linux flavors.

build_ocean.sh and build_simple.sh are converted to interactive scripts to enable alias loading for uses who need to alias specific binaries e.g. clang

When there isn't enough arguments to execute a command, a command is invalid or the user tries to access a help section for a command without specifying en environment which isn't defined a global HELP_MESSAGE is displayed instead of a massive error which can confuse new users.

It would be nice to rewrite the cli to typer https://typer.tiangolo.com/ to enable autocomplete and nicer documentation. Is this something that you think would be useful?

I really like the feeling of seeing the neural net train itself only from playing with itself. I wanted to write the environment for tic tac toe but I can see that there is already an open pr for this. Can you suggest a similarly difficult environment to contribute, potentially with a guide to writing one, similar to the readme?

[leanke]

https://puffer.ai/docs.html heres a link to the docs 😉

[Anakin100100]

I know, I read them. It's just that the setup instructions and step by step instructions on training, exporting and using an env are missing. It could be a good idea to copy this to the docs and enhance the guide for writing a new env because this one is somewhat confusing because it doesn't provide a guide on writing a new env it just points to a quite simple one and then shows how to perform some tasks after it has already been written without explaining the process of designing the env, agent evaluation etc.

Having the end to end walkthrough in readme is imho important because a lot of people find this first so having it in both places would be useful.

[elevatorguy]

Pufferlib aside, using dash by default - instead of bash - and running a compile or build script that relies upon .bashrc, for example oneapi initialization; thanks.