Function qip::iterators::sum_for_op_cols

pub fn sum_for_op_cols<P: Precision, F: Fn((u64, Complex<P>)) -> Complex<P>>(
    nindices: u64, 
    row: u64, 
    op: &PrecisionUnitaryOp<P>, 
    f: F
) -> Complex<P>

Using the function f which maps from a column and row to a complex value for the op matrix, sums for all nonzero entries for a given op more efficiently than trying each column between 0 and 2^nindices. This really needs to be cleaned up, but runs in a tight loop. This makes it hard since Box is unfeasible and the iterator types aren't the same size.