🚀[FEA]: Making scoring resumable

[nbren12]

Is this a new feature, an improvement, or a change to existing functionality?

New Feature

How would you describe the priority of this feature request

Critical (currently preventing usage)

Please provide a clear description of problem you would like to solve.

Scoring large models is an expensive workload, and sometimes cannot be completed in a single session. It would be nice to make both lagged ensembles and medium range resumable.

Alternatively could add a --n-shard --shard flags to the medium range inference scripts to make the problem size small enough to complete in one session.

This is blocking graphcast scoring.

Describe any alternatives you have considered

No response