Tried tuning ECS

[ptrscll]

I tried including ECS as a tunable parameter while optimizing NMSE accounting for uncertainty in temperature data. Results are as follows:

• For both the regular and big box constraints, optimal S was 5, leading to unrealistic future warming.
• Temperature MSE decreased from previous runs.
• CO2 MSE was still very high due to using NMSE, but it was very slightly lower than in the equivalent experiments done with a fixed S.

A plot showing the results of these initial runs is given below:

For next steps, I might try to re-weight the error function to better balance temperature and CO2 error, but even when including S as a tunable parameter, it doesn't look like too much improvement in temperature results is possible (without incorporating another parameter).

This PR is related to #8