Integrate the well training approaches with the current FNO example for reaction/advection-diffusion

[sgreenbury]

This aims to update the current example notebook using a version of the spatiotemporal emulator / dataset / training with one that is also compatible with the well datasets.

So you can substitute the different datasets out at the beginning, change the args of the emulator/dataset/formatters and it will run.

Aim: use the trainer, models and formatters from the well.

Future: implement new data formatter to enable time for e.g be used.

[sgreenbury]

Close not planned (see alan-turing-institute/auto-z#16)