Add initial support for running METplus on precip observations at NCI

### What problem does your feature request solve?

Add initial METplus support for precip observations

There are multiple possible source data types, both point and gridded, we want to be able to compare all of them.

### Describe the solution you'd like

In rose-suite.conf enable Precipitation verification with

```
METPLUS_VERIFIY_PRECIP=True
```

The user needs to be able to supply what the rainfall variable is, and what its accumulation period is.

Variables can be:
- `stratiform_rainfall_amount` for UM
- `ls_precip` for LFRic

Until there's a normalised way to refer to the same variable in different models let's just allow the user to specify what field to compare and the MET-style accumulation period in rose-suite.conf:

```
METPLUS_PRECIP_FCST_VARIABLE='stratiform_rainfall_amount'
METPLUS_PRECIP_ACCUM='A3'
```

When `METPLUS_VERIFY_PRECIP` is set, verify against point and/or gridded data depending on the state of `RUN_METPLUS_POINT_STAT` and `RUN_METPLUS_GRID_STAT`.

This should be a generic enough interface that other sites can plug in their own METplus configs for comparing against different precipitation data sources.

Plotting the results with verpy is left to a future step, this ticket is only concerned with the user interface of running METplus.

### Describe alternatives you've considered

Precipitation data is special-cased from the generic point-stat and grid-stat configs to be able to choose different data paths based on the accumulation period being used.

If precipitation verification was always enabled then the user would need to always remember to set the variable name appropriate for the model when setting up CSET, which seems like a burden if they're not needing to look at precip values.

Comparing UM and LFRic models in a single CSET run is out of scope here, but you could imagine allowing METPLUS_PRECIP_FCST_VARIABLE to optionally take a list with a different variable for each input model.