Support interpolation on dask arrays

[bouweandela]

First attempt at supporting dask arrays.

See #47 and ESMValGroup/ESMValCore#35.

[bouweandela]

@bjlittle If you look at the profile_stratify.ipynb notebook, you'll see that the problem originally was that interpolating to a single target level was very much slower on a dask array than on a numpy array. However, when I run the same code now the problem seems to have disappeared, so I think the approach here is viable.

[bjlittle]

@bouweandela Awesome, thanks! 🥳

I'm simply going to merge this PR "as-is", even though the CI is failing.

The infra-structure of python-stratify needs quite a bit of love as it's sadly gone stale, so I'm going to take care of that as a separate concern.

Any changes required to tweak your PR contribution can be addressed as an immediate follow-up before making this lazy feature available in a forth-coming release - I'm happy to do that on your behalf 👍

The tests are passing locally, and that's good enough for me given the current state of affairs to unblock this PR.