Fix cblas_nrm2 for complex types

[lokeshh]

Pass correct parameters to nrm2_complex_helper

Fixes #389

[v0dro]

Could you write an accompanying test that proves that the hand-coded version returns the same answer as the BLAS version? Preferably use the example raised in the issue.

Also rebase your commits later.

[lokeshh]

I do not know what the function should do in case there's a complex number in the vector. For example if the vector is [1, 1+2i, 3], what should be the output?

From the BLAS version, it seems that it should be (12 + 12 + 32 + 2**2)0.5. I was wondering if there's some documentation just to be sure, so no corner cases are left.

[translunar]

@lokeshh It is defined here, in the comments: http://for2html.sourceforge.net/BLAS/dznrm2.html

#  DZNRM2 returns the euclidean norm of a vector via the function
#  name, so that
#
#     DZNRM2 := sqrt( conjg( x' )*x )

[lokeshh]

Does this look good?