provide quick overview of the main principle of the library

[jerabaul29]

After discussing with some colleagues who I recommended to look at this repo, they said that what took them a bit of time was to understand what coupling strategy was used, and that it may be a good idea to make it clear in the readme with a paragraph / maybe a workflow figure.

I.e. making it clear that the strategy this code uses is that:

• at each new action, a new OpenFOAM simulation is started from the last dumped fields
• this new simulation runs until the end of the action time, then the fields and dumped and the information for the DRL agent are dumped too

I.e., there is no "direct online coupling" through sockets or MPI between the OpenFOAM simulation and the DRL, but a new OpenFOAM simulation is restored from the previous dumped state at each action.