Track and write out spatial map of model fit failures

[tsalo]

Summary

Currently we just let users know how many model fit failures there are in the nonlinear T2*/S0 estimation step, and patch in the log-linear estimates into failing voxels, but I think it would be useful to write out a boolean map indicating if the model was successful or not.

[tsalo]

I have found that voxels with model fit failures end up having terrible loglinear T2* estimates, so one thing I'd like to do that builds off this is to give users the option to interpolate their T2* and S0 values over these failed voxels.

[tsalo]

Got the go-ahead to do this in the devs call today.