Remove AbstractIMF?

[cgarling]

Since the PowerLawIMF and LogNormalIMF constructors just create normal truncated distributions and not AbstractIMFs, I'm not sure the abstract type actually helps anything. We're not defining any special methods for AbstractIMFs, so that doesn't support the addition of the abstract type either.

[cgarling]

We might use AbstractIMF later if we decide to add normalized volumetric IMFs or something like that ... doesn't seem to be hurting anything for now.