Support SemanticPointer operations with numpy scalars

[arvoelke]

The expression:

spa.SemanticPointer(np.zeros(1)) * np.float64(1)

produces the error:

~/git/spa/nengo_spa/semantic_pointer.py in __mul__(self, other)
    227         If multiplied by a scalar, we do normal multiplication.
    228         """
--> 229         return self._mul(other, swap=False)
    230 
    231     def __rmul__(self, other):

~/git/spa/nengo_spa/semantic_pointer.py in _mul(self, other, swap)
    239         if is_array(other):
    240             raise TypeError(
--> 241                 "Multiplication of Semantic Pointers with arrays in not " "allowed."
    242             )
    243         elif is_number(other):

TypeError: Multiplication of Semantic Pointers with arrays in not allowed.

This is relevant because functions such as np.mean, np.sum, and np.linalg.norm, can all return numpy scalar types. For example,

sp = spa.SemanticPointer(np.zeros(1))
sp *= 1 / np.linalg.norm(sp.v)

produces the same error (note I am using multiplication instead of division because of #249, and yes I am aware of sp.normalized() -- this is just an example).

As a work-around, one can use .item() to convert the scalar to a native Python type, like so:

spa.SemanticPointer(np.zeros(1)) * np.float64(1).item()