Sex should not be an enum

[driusan]

Sex is currently defined in the candidate table as an enum with the values enum('Male','Female','Other'). This is to align with the possible DICOM values, but is archaic from a scientific perspective and non-inclusive.

We should normalize the possible sex values into another table, rather than using an enum so that they can be customized per LORIS instance according to the study design of the project.