fixed bug in decompose_dipole, eegmegcalc.py

[solveignaess]

When implementing #142, I accidentally left in a bug which would give potentials with opposite sign for dipole locations x<0/ y<0 /z<0. =/

Fixed bug and added test.

[coveralls]

Coverage increased (+0.008%) to 97.57% when pulling c3298b2 on solveignaess:4S_diploc into deb24ce on LFPy:master.

[espenhgn]

Thanks, will merge when the tests pass Travis.

[espenhgn]

Hi @solveignaess. I'll merge this now, but this code could be made a bit prettier some time by turning z into a 2D array, something like:

z_ = np.expand_dims(z, -1)
...
p_rad = np.dot(np.dot(p, z_), z_.T)
# produces identical output as:
# p_rad = np.dot(p, z).reshape((len(p), 1)) * z.reshape((1, z.size))