Explore better dependency management strategy for data-analyses

[raebot]

We'd like to have cleaner delineation of dependencies between any given analysis project and other analysis projects as well as their package dependencies. This would make it more intuitive to get started on any project and easily know what all is needed to run it. It would also make it easier to break projects into their own repos, if desired.

Do we still want to maintain a small, core set of dependencies that the vast majority of analysis projects use? What's the best way to do this? Currently this is done via the jupyter-singleuser image and the calitp-data-analysis package, but folks reported that these packages are sometimes outdated for what they need and they end up having to do their own installs anyway. Also there are probably more packages in these than are strictly needed.

Do we want to switch to uv? Benefits of uv over Poetry or plain pip

• very fast package installs because it's written in Rust and makes more use of caching
• can manage Python version in addition to package versions
• virtual environments