Topic: Bugfix for Non-Linear problems group

I have committed fix for a bug that was related to handling sparse matrices from FuncDesigner AD in openopt - there was a check
if r.size == 1
made for elements of type numpy.ndarray, so if r is scipy.sparse than r means number of non-zero elements instead of number of all elements (and this is one of serious drawbacks of using scipy.sparse). I had missed the place in code when implemented handling of scipy.sparse matrices. Another one: asfarray(sparse_matrix) yields array of single element with data type "object".

If you don't deal with FuncDesigner but pass scipy.sparse matrices as user-supplied derivatives to OpenOpt, the bug with r.size can affect as well.