[FR] Add support for custom dataloaders

[james-alvey-42]

Is your feature request related to a problem? Please describe.
Add support for custom pytorch dataloaders so that not entirely dependent on zarr stores

Describe the solution you'd like
Modularise the training function so that it can take in a zarrstore or custom dataloader from the user

[james-alvey-42]

Not necessarily about writing lots of custom ones ourselves, just support for any user wanting to implement their own - but need to give an example of what the dataloader needs to look like