Is your feature request related to a problem? Please describe.

The basic system-level test suite added in PR #38 has a family of tests for dataset creation. However, it currently only tests if dataset creation crashes and if a zarr dataset exists in the expected location (and if training tasks that are based on the created datasets crash). We should add a couple of checks that verify if the dataset looks as expected.

Describe the solution you'd like

We could compare to a reference dataset, or we could make checks configurable in the config folder for the dataset test such that only relevant checks for the dataset are run. Any checks should be added to the task that runs the dataset test (which currently consists of running the anemoi-datasets create command)

Describe alternatives you've considered

No response

Additional context

No response

Organisation

No response