Add complex number support to linalg.vector_norm

[kgryte]

This PR

• adds complex number support to linalg.vector_norm. The main substantive change is updating the input and output data types. For both real- and complex-valued inputs, norms (apart from ord=0) involve computing the absolute value, so no changes are necessary in norm definitions to accommodate complex numbers.
• requires that the output data type always be a real-valued floating-point data type.

[rgommers]

This looks pretty straightforward, in it goes. Thanks @kgryte