Impose/assume specific units when reading profiles.

[pazathoth]

Currently, the subroutines read_profiles and compute_auxiliary_profiles apply a heuristic to determine whether the independent variable is given in terms of effective radius or the normalized poloidal flux. This choice should be fixed, i.e., it's the user's responsibility to provide the data in the expected format. Alternatively, switches could be added to specify the choice explicitly.