Automate figure generation for docs

Right now, figures for tutorials and documentation have to be manually generated when we make a release. This often leads to out of date figures, and a fair amount of pain for otherwise small style changes. It would be good if this could be part of an automatic build process.

For the sphinx docs, we could probably do this on the readthedocs servers since we have extra capacity.

For notebooks, this could take some consideration. Some notebooks deal with large data that we don't want to reprocess frequently. A good step could just be some simple scripting around automatically running the notebooks.

Start by reviewing the Sphinx documentation build and the notebooks mentioned in the issue, including how releases currently generate figures. Define how automatic generation should handle large-data notebooks and Read the Docs capacity; done means release builds produce up-to-date figures without manual generation.