Add support for running tasks on a Dask cluster

[bouweandela]

Since iris 3.6, it is possible to use Dask distributed with iris. This is a great new feature that makes it possible to implement distributed computing (e.g. running a single recipe on multiple cluster nodes) in ESMValCore and ESMValTool. See #1714 for an example implementation. I would like to add this feature.

[bouweandela]

After gaining some more experience with Dask I'm not sure if this is a good idea because

• diagnostic tasks are likely to use more memory than workers have
• for preprocessing tasks, metadata handling may also exceed the amount of memory that is available on a Dask worker, though this could probably be fixed
• for preprocessing tasks a more fine-grained approach would parallelize much better