Fix unit string representation

[mstimberg]

As @romainbrette noticed, the string representation of some units were not correct in the sense that evaluating them in Python did not result in the same unit – this is obviously bad, in particular when storing units/equations in text and importing them later (as Romain did). The problem concerned units raised to powers more than once, e.g.:

>>> (second ** 2) ** -1  # or 1/second**2 which will be interpreted in the same way
second ** 2 ** -1

Due to Python's operator precedence, this will be interpreted as second ** (2 ** -1), i.e. second ** 0.5 instead of second ** -2 as it should be. The simple fix in this PR is to consider units raised to powers as "compound units", which will lead to the use of parentheses:

>>> (second ** 2) ** -1
(second ** 2) ** -1

Note that this changes the output also in situations that do not need correction, e.g. amp / cm ** 2 will now be represented as amp / (cm ** 2). I'd rather keep the fix as simple as it is though, the new output might even be helpful to avoid confusion.

While fixing this I came across another minor issue which is now fixed as well: the function is_scalar_type was supposed to return True for a scalar that "can be interpreted as a dimensionless Quantity", but it actually only checked if it was a scalar, not whether it was dimensionless. Since this function was used to check whether an exponent of a unit was useable, it did give a somewhat confusing error message for nonsensical operations like second**second:

...
brian2.units.fundamentalunits.DimensionMismatchError: Cannot calculate 10.0 ** 0. s, the exponent has to be dimensionless (unit is s).

With the fix, the error message is more meaningful

brian2.units.fundamentalunits.DimensionMismatchError: Cannot calculate s ** s, the exponent has to be dimensionless (unit is s).