Add juv venv to support exporting explicit notebook enviroments

[manzt]

Alternative to #62

Some editors and enviroments are missing the benefits from juv standalone notebooks, simply because we don't make it easy to use the specified virutal enviroment outside of spawning a Jupyter runtime with juv run.

While we typically try to hide virtual environments, they remain crucial for ecosystem like language servers, autocomplete, and intellisense, and exposing some way to quickly make a notebok enviroment is the easiest way to bring juv standalone notebooks to unknown tools.

To improve compatibility with more editors and tools, this PR introduces:

juv venv --from=Untitled.ipynb

This command creates a virtual environment with the dependencies and locking behavior expected from juv, and adds ipykernel for running the notebook. The resulting environment can be selected in an editor like VS Code.

To create a venv with only the locked dependencies, add the --no-kernel flag:

juv venv --from=Untitled.ipynb --no-kernel

Alternatives

While editor-specific integrations for juv would be ideal (and still welcome!), they require significant effort. I anticipate there will be opportunities to piggyback on whatever editors do to support uv standalone scripts in the future.

In the meantime, this approach provides a decent escape hatch. I think it's a reasonable tradeoff for individuals that want to use their favorite editor with notebook support and also read/write standalone notebooks.

Note

It's worth noting that these commands are basically sugar for composing some lesser known juv and uv commands:

uv env
juv export Untitled.ipynb | uv pip install -r -

or, when including ipykernel:

uv env
(juv export Untitled.ipynb; echo ipykernel) | uv pip install -r -

However, I think having a simple high-level API is ideal. It also means one does not need to have uv installed (even though they should!)