Bug in LFW extension

The scores of an experiment using LFW gives me this output, which is obviously incorrect:

13 Enos_Slaughter Enos_Slaughter/Enos_Slaughter_0001 -0.32567173
13 Juan_Carlos_Ortega Juan_Carlos_Ortega/Juan_Carlos_Ortega_0001 -0.33916610
13 Marlon_Devonish Marlon_Devonish/Marlon_Devonish_0001 -0.31824485
18 Abdel_Nasser_Assidi Abdel_Nasser_Assidi/Abdel_Nasser_Assidi_0002 -0.04248074
51 Bruce_Paltrow Bruce_Paltrow/Bruce_Paltrow_0001 -0.40190915
51 Howard_Smith Howard_Smith/Howard_Smith_0002 -0.37437308
51 Jan_De_Bont Jan_De_Bont/Jan_De_Bont_0001 -0.41860976
51 Jim_Nochols Jim_Nochols/Jim_Nochols_0001 -0.27871283
51 Oleg_Romantsev Oleg_Romantsev/Oleg_Romantsev_0001 -0.35602057

For most of the databases the model_id is equals to client_id, which is not the case for this dataset.

bob.bio.face.database.LFWBioDatabase should reimplement the method bob.bio.face.database.BioDatabase.client_id_from_model_id and fix this.