Treating categories not as numerical variables

[jvsoest]

Currently the coxph algorithm only accepts numerical variables.
Please include support for categorical variables as well (as e.g. a tumor stage distance between T0 and T1 is possibly different from T2 to T3).

This could be solved by detecting if factor columns are in the input variables, and make boolean columns (containing a numerical 0 or 1) for every factor. The first factor in the list can then function as the baseline (0 for all boolean columns).