[compliance] walker_lake

[maximlt]

Main changes as I had some issues getting the example to run:

• No more intake: intake + intake-xarray was raising some errors, replaced by anaconda-project downloads
• rioxarray used to open the files
• Bigger chunks: even with bigger chunks I find the example slow with Dask, the data fits in the memory, yet it's a good example to show how to use Dask (maybe there are other knobs to tweak, maybe some performance optimisations in our HoloViz toolset)
• No more resampling/regridding with Datashader: After the package upgrade, the mean/sum plots were showing very little data (I'm not sure why), enough though to skip the regridding branch. It worked better when regridding was forced. I imagine that geo users wouldn't regrid their data with datashader+geoviews. I tried rioxarray.project_match which worked great but is not Dask compatible (it just loads the whole dataset in memory). xarray.interp_like worked perfectly and gave the same results as rioxarray.project_match.

Note that this is not a full update of this example (a la NF SDG), it's just to adapt it to the (not so) new infra (e.g. more validation checks) like we did a year ago for most of the examples.

[github-actions]

Your changes were successfully integrated in the dev site, make sure to review the pages of the projects you touched before merging this PR.