simplify get_bitinformation

[observingClouds]

fixes eventually #261

[thodson-usgs]

Any plan to continue on this? Trying to decide whether and how best to resume: zarr-developers/numcodecs#503

I had advocated for a pure numpy implementation of the bitinfo algorithm that would eventually be included in xbitinfo and from there imported into numcodec, but perhaps that was jumping ahead... An alternative would be to just import the xarray-based implementation from xbitinfo, use that in the codec, but make the module optional until we have a pure numpy implementation: check if xarray is installed and throw and error if it isn't.

On the xbitinfo end. Ideally, we only maintain a single version of the algorithm. I think a numpy implementation might replace the lower-level implementation, where xarray doesn't seem as well suited, but there would still be the same high-level xarray interface wrapping it.

[observingClouds]

Thanks @thodson-usgs, yes I do plan to continue this PR, hopefully I come back to this within the next weeks.