Use fsspec and/or pathy to provide a unified interface to GCP, AWS, and local filesystems

[JackKelly]

At the moment, our code implements different functions to interact with data on GCP vs the local filesystem. This is exactly the problem that fsspec solves: fsspec provides a unified interface to cloud and local storage. If we use fsspec then we can probably throw away quite a lot of our filesystem functions :)

Pathy also looks very useful.

Before starting work on this issue, we should wait for #152 to be merged into main (UPDATE: #152 is now merged into main!)

We can probably remove the cloud field in the yaml config (in model.py) after making this change, because the code will be able to infer the compute platform from the protocol prefix on the paths (e.g. gs:// for Google Cloud).

[JackKelly]

I'm making a start on this in #152 (only where necessary to make it easy to write to local filesystem). I'll leave it to a later PR to complete this task.

[JackKelly]

A word of warning: Pathy.exists() seems buggy to me: Pathy('gs://bucket/foo/bar.txt').exists() sometimes hangs. Pathy('/local/file.txt').exists() sometimes returns the wrong answer (although maybe we're supposed to include a protocol prefix like 'file://local/file.txt'?) We should explore this more and, if this really does turn out to be a bug, we should report it to the Pathy project.

That said, using Pathy as a prettier way to join paths works really well. e.g. Pathy('gs://bucket') / 'foo' / 'bar.txt'