allow density/normed argument

Thank you for your cool package!

`numpy.histogram` and `numpy.histogram2d` support the `normed` argument, which divides counts by the number of samples and the bin width/area. It would be great if fast-histogram could support this as well.

Compare fast-histogram's public 1D and 2D histogram entry points with numpy.histogram and numpy.histogram2d, focusing on their normed argument behavior. Confirm the expected normalization by sample count and bin width or area, then add coverage for both dimensions; done means the package accepts the argument and matches NumPy's results.

Written by the indexing model from the issue text.