Pow.as_base_exp vs I.as_base_exp; (-1)**e -> I**(2*e)

[smichr]

>>> I.as_base_exp()
(-1, 1/2)
>>> sqrt(I).as_base_exp()
(I, 1/2) <-- could be (-1, 1/4)

[smichr]

I think we can be more canonical about how I is treated:

>>> I**(1/3)*(-1)**(1/4)
(-1)**(1/4)*I**(1/3)  <--- could be (-1)**(7/12) or I*(-1)**(1/12) or I**(5/6)

Rather than keep powers of (-1) separate perhaps (-1)**e should always be written as I**(2*e)

[jksuom]

perhaps (-1)e should always be written as I(2*e)

I'm not sure about this. That would introduce I in almost all expressions.