June '23 with XIOS

[joewallwork]

Being able to run the June '23 configuration with 4 MPI ranks with XIOS would make for a great integration test.

[joewallwork]

The current status is that I can run up to getting an error

Xios: Field 'cice' found in both the XiosInput and XiosOutput config sections

This makes #1007 is a priority.

[joewallwork]

The above issue was resolved in #1017.

The next issue was then that it needs to be possible to read a field as HField and interpret it as DGField in the model - see #1034, which should be fixed by #1037.

[joewallwork]

The next issue is to deduce the field types for XIOS. The current draft approach can be found in #1043.

Running the June '23 config on this branch, I hit the error

ParaGridIO::dumpModelState: field tsurf is not configured as a restart.

It appears that the tsurf variable is in the model state and is passed to dumpModelState but it hasn't been configured for writing as a restart.

[joewallwork]

The issue seems to be that physics adds tsurf to the model state no matter what you do. So we need to allow getModelState and dumpModelState to skip fields it sees in the model state that aren't configured for XIOS I/O.

[joewallwork]

The latest version of the branch issue982_xios-june23 with a very trimmed-back config file produces restart files at each timestep! It gets as far as finalization and then hits a std::out_of_range error.