Investigate issues found when running models

User Story: For interval 12 (and previous intervals) the stimulation done resulted in various errors that meant the model runs sometimes did not complete.

1. Deserialization error where marshmallow thinks that nans are found

2. Sometimes no forecast is returned to ramsis. I think this is the result of something going wrong in the model and not completing

3. error returned to ramsis but data attached to model run. Why is there an error in this case?

4. Running two forecasts that are the same (same time and data) with 3 scenarios each resulted in a different scenario failing each time (and two success).

Acceptance criteria: should be able to run a series of forecasts from interval 12 with no errors from models (assuming it is something not to do with model itself)