Catch exceptions raised when optimizer fails due to low quality data

[laurencejackson]

The Philips MRS data passes through a 3rd party function from the suspect library called “correct_frequency_and_phase”, from what I understand this function performs a registration of the frequency and phase of each MRS measurement stored in the dynamics. T

The method chosen for this correction is currently RATS (Robust Alignment to a Target Spectrum) which comes with a caveat from the author of the function that it has problems with getting stuck in local minima. The optimizer used (Brent) uses a bracket for defining the search space, this is by default a course search at +/- 20Hz, then a refined search.

When data quality is poor this correction function fails and no exception is raised.

In addition, we should add a fallback method that registers the data using a less sophisticated method in cases where data quality is low. e.g. Time-domain spectral registration