feat: copy zarr chuncks lazily to ssd

[b8raoult]

Description

A Zarr store that copy chunk lazily to an other filesystem (ssd based), so that the second epoch of training is faster.

Example, scanning aifs-ea-an-oper-0001-mars-o96-1979-2022-6h-v6 twice:

CopyToSSDStore: using temporary directory /.../ssd1/tmpdirs/.../anemoi-datasets-ssd-prv26ojg
Pass 1: 100%|████████████████████████████████████████████████████████████| 64284/64284 [1:07:32<00:00, 15.86it/s]
Pass 1 took 4052.78 seconds
Pass 2: 100%|██████████████████████████████████████████████████████████████| 64284/64284 [15:16<00:00, 70.15it/s]
Pass 2 took 916.36 seconds
CopyToSSDStore: total size copied: 578.7 GiB
CopyToSSDStore: copied 64,287 objects, reused 64,284 objects

First pass (From disk + copy to SSD) : 1h
Second pass (from SSD): 15 minutes

What problem does this change solve?

What issue or task does this change relate to?

Additional notes

As a contributor to the Anemoi framework, please ensure that your changes include unit tests, updates to any affected dependencies and documentation, and have been tested in a parallel setting (i.e., with multiple GPUs). As a reviewer, you are also responsible for verifying these aspects and requesting changes if they are not adequately addressed. For guidelines about those please refer to https://anemoi.readthedocs.io/en/latest/

By opening this pull request, I affirm that all authors agree to the Contributor License Agreement.