Treatment for multiple instances of same named cube

### What problem does your feature request solve?

Changes to plot multiple model outputs on same plot (e.g. histogram, timeseries, profiles) mean that recipes are run with
`operator: read.read_cubes`
This is required and desirable for multi-line plotting.
However, this can lead to CSET failures where multiple instances of the same cube are read in from a file. I think this happens where for example a time-mean and instantaneous output are both provided in a set of file streams (these need not even be in same file if multiple files read in, e.g. LFRic diagnostics and time_proc files).

### Describe the solution you'd like

Where possible, a check for cube.cell_methods should be made where multiple cubes meet a given constraint. This could be restricted to only a few 'known' variables to prevent missing required variables by always forcing time-mean inputs, or only triggered if >1 cube returned.

### Describe alternatives you've considered

No implementation files or tests are named. Start by tracing the read.read_cubes operator and how cube.cell_methods is handled when multiple cubes satisfy a constraint. Define the expected selection behavior for time-mean and instantaneous outputs, then add coverage for multiple same-named cubes and confirm existing multi-line plotting behavior is preserved.

Written by the indexing model from the issue text.