When lfric_atm starts at timestep zero from an lfric2lfric dump file, it should not try to read random_seed etc

### Are there any linked Issues or Pull Requests?

_No response_

### Brief description

lfric_atm checkpoints some non-field arrays such as random_seed, temperature_correction_rate and several arrays that influence stochastic physics. Checkpointing these items is required to support continuation runs.

The checkpointing is set up in gungho_model_mod.F90

Most likely, it makes sense always to start a new run (one that starts with timestep zero) with these arrays initialised to fresh values.

The current setup is problematic for runs that are initialised from lfric2lfric whose dumps will not include these arrays.

### Further details of the issue.

_No response_