Promote land masking as wider option in CSET

### What problem does your feature request solve?
Currently we have an option to produce timeseries with a surface field and mask this (with land if present). I don't think this supports aggregation, or pressure level variables. There may be cases where we also want to compute a histogram out of masked data (land vs sea). This has been flagged by @mo-rwjones and Seren.

### Describe the solution you'd like
We could have some option higher up in the GUI, such as subarea cutout stage, where we ask a user if they want to mask all data in subsequent analysis. Some things might break with np.nans present (e.g. power spectra). We would also have to relax the constraints to ensure that land_mask is allowed to load.

### Describe alternatives you've considered
Masking data outside of CSET prior to running (could do, but seems quite cumbersome when we have the functionality in CSET.
An option to land or sea mask for every option in CSET, but would make GUI overly verbose.

Tagging @daflack as original developer to add this functionality in case thoughts

Start by locating the GUI's subarea cutout stage and the constraints that control loading land_mask. Trace how masking would propagate into aggregation, pressure-level variables, histograms, and analyses affected by np.nan values. Done means a higher-level land-or-sea masking option is available and its downstream behavior is covered for the cases described.

Written by the indexing model from the issue text.