spectrum_per_label() results are not consistent with other laplace_beltrami.py functions

[binarybottle]

for label 22, rather than getting:

([[6.3469513010430304e-18,
   0.0005178862383467463,
   0.0017434911095630772,
   0.003667561767487686,
   0.005429017880363784,
   0.006309346984678924]],
 [22])

as in the other functions, i'm getting:

eigsh() failed. Now try lobpcg. Warning: lobpcg can produce different results from Reuter (2006) shapeDNA-tria software. Compute linear FEM Laplace-Beltrami spectrum Out[4]: ([[9.9512415855027014e-06, 0.037430186764067855, 0.040538097886371673, 0.044689929817993072, 0.051117433202677504, 0.060568241779909528]], [22])

[binarybottle]

Fixed by reindexing faces.