Provide possible solutions to unexpected variable splitting

[neishm]

There are many scenarios where an FST variable may be split into multiple names (with extra suffixes) when converting to netCDF or creating an xarray object in Python. This may be indicative that extra parameters need to be passed to the module, but the specific option may be buried deep in the documentation. Should provide some helpful suggestions as part of the warning messages, E.g.,

• When multiple typvars are found (both analysis (A) and forecast (P)), then suggest --ignore-typvar to keep things together.
• When multiple etikets are found, this could either be because the version of the operational system changes within the list of input files, or this is an ensemble system. Suggest either --ignore-etiket or --ensembles depending on the context.
• If multiple vertical coordinates are found for the same variable (such as model hybrid levels and a diagnostic height level), suggest --ignore-diag-level, --only-diag-level, or --diag-as-model-level. (note: this hint is already there, but could be factored out into a better hint system).
• If the same variable occurs with multiple horizontal domains, suggest maybe --crop-to-smallest-grid if the grids happen to be colocated (increase / decrease in grid domain within a data product?)
• Probably many other corner cases that could be added here as they are encountered.

For the Python interface, need to modify the message for the corresponding Python parameters (e.g. suggest ignore_typvar=True, etc.)