Update `make_ww3_*_file` workers and their place in the automation workflow

[douglatornell]

• [x] Add mpi hosts file to wave forecasts configuration

  Enables wwatch3 to be run on different VMs to those where NEMO is run. This is a step towards running wwatch3 nowcast/forecast runs concurrently with nowcast-green.

• [x] Add chunks and drop_variables parameters into xarray.open_mfdataset() calls

  The hope is that these changes will resolve (or at least reduce the frequency of) the issue of one or both of the make_ww3* workers stalling during the ds.to_netcdf() step in which dask is computing the task graph.

• [x] Move make_ww3_*_file worker launches to after NEMO forecast run so that wwatch3 nowcast and forecast runs are concurrent with NEMO nowcast-green run

[douglatornell]

Launch of make_ww3_*_file workers after NEMO forecast run was successful in production on 15 & 16 Nov.

[douglatornell]

Stalling of one or both of the make_ww3* workers during the ds.to_netcdf() step in which dask is computing the task graph remains unresolved. I'm out of ideas for how to fix that via changes to the workers. So, it's time to merge this.