change transforms.py to represent "scalars" as zero-D arrays

[CrepeGoat]

originally discussed in #44 (comment).

The transforms.py code requires if-else logic to specially handle inputs & internal variables, depending on whether they're numpy arrays or scalar values. This special-case logic could, in most cases, be removed by representing "scalars" as zero-dimensional arrays, which would remove excess complexity from the system.

One example of special-case logic occurs on this line here:

            if scalar:
                out = self.function(values)
            else:
                out[:] = self.function(values)

By representing the scalars as 0D-arrays, ^this could be generically refactored to v this:

            out[...] = self.function(values)

Implementing this change will likely break a handful of other dependent pieces of code, since 0D-arrays don't exactly behave the same as true scalars in all situations; this should be investigated in more detail.