initialized wk before adapt loop

[yuvlyg]

pmtm with "adapt" method crashes if data is a constant vector
Loop is not entered (since convergence criterion is already met), so wk should be initialized before it.

[coveralls]

Coverage decreased (-0.0009%) to 97.505% when pulling 4ef0307 on yuvlyg:master into a02bf9d on cokelaer:master.

[cokelaer]

Sorry but I am probably missing something here. I tried with a constant vector as input but it seems to work:

from spectrum import mtm
m = mtm.pmtm([1,1,1,1,1,1,1,1])

and it does not crash. insteqd it returns valid vectors. Do you have an example of a crash that I can reproduce. thanks

[yuvlyg]

My bad, it happens with zeros, not any constant vector:

from spectrum import mtm
import numpy as np
m = mtm.pmtm(np.zeros(100), NW=4)
# we get:
# UnboundLocalError: local variable 'wk' referenced before assignment