Make notebook support optional

[wimglenn]

For 1.4 the dependency tree of removestar looked like

removestar<=1.4
└── pyflakes

After https://github.com/asmeurer/removestar/pull/31 it looks like

removestar==1.5
├── nbconvert
│   ├── beautifulsoup4
│   │   └── soupsieve>1.2
│   ├── bleach!=5.0.0
│   │   ├── six>=1.9.0
│   │   └── webencodings
│   ├── defusedxml
│   ├── jinja2>=3.0
│   │   └── MarkupSafe>=2.0
│   ├── jupyter-core>=4.7
│   │   ├── platformdirs>=2.5
│   │   └── traitlets>=5.3
│   ├── jupyterlab-pygments
│   ├── markupsafe>=2.0
│   ├── mistune<4,>=2.0.3
│   ├── nbclient>=0.5.0
│   │   ├── jupyter-client>=6.1.12
│   │   │   ├── jupyter-core!=5.0.*,>=4.12
│   │   │   │   ├── platformdirs>=2.5
│   │   │   │   └── traitlets>=5.3
│   │   │   ├── python-dateutil>=2.8.2
│   │   │   │   └── six>=1.5
│   │   │   ├── pyzmq>=23.0
│   │   │   ├── tornado>=6.2
│   │   │   └── traitlets>=5.3
│   │   ├── jupyter-core!=5.0.*,>=4.12
│   │   │   ├── platformdirs>=2.5
│   │   │   └── traitlets>=5.3
│   │   ├── nbformat>=5.1
│   │   │   ├── fastjsonschema
│   │   │   ├── jsonschema>=2.6
│   │   │   │   ├── attrs>=22.2.0
│   │   │   │   ├── jsonschema-specifications>=2023.03.6
│   │   │   │   │   └── referencing>=0.28.0
│   │   │   │   │       ├── attrs>=22.2.0
│   │   │   │   │       └── rpds-py>=0.7.0
│   │   │   │   ├── referencing>=0.28.4
│   │   │   │   │   ├── attrs>=22.2.0
│   │   │   │   │   └── rpds-py>=0.7.0
│   │   │   │   └── rpds-py>=0.7.1
│   │   │   ├── jupyter-core
│   │   │   │   ├── platformdirs>=2.5
│   │   │   │   └── traitlets>=5.3
│   │   │   └── traitlets>=5.1
│   │   └── traitlets>=5.4
│   ├── nbformat>=5.7
│   │   ├── fastjsonschema
│   │   ├── jsonschema>=2.6
│   │   │   ├── attrs>=22.2.0
│   │   │   ├── jsonschema-specifications>=2023.03.6
│   │   │   │   └── referencing>=0.28.0
│   │   │   │       ├── attrs>=22.2.0
│   │   │   │       └── rpds-py>=0.7.0
│   │   │   ├── referencing>=0.28.4
│   │   │   │   ├── attrs>=22.2.0
│   │   │   │   └── rpds-py>=0.7.0
│   │   │   └── rpds-py>=0.7.1
│   │   ├── jupyter-core
│   │   │   ├── platformdirs>=2.5
│   │   │   └── traitlets>=5.3
│   │   └── traitlets>=5.1
│   ├── packaging
│   ├── pandocfilters>=1.4.1
│   ├── pygments>=2.4.1
│   ├── tinycss2
│   │   └── webencodings>=0.4
│   └── traitlets>=5.1
├── nbformat
│   ├── fastjsonschema
│   ├── jsonschema>=2.6
│   │   ├── attrs>=22.2.0
│   │   ├── jsonschema-specifications>=2023.03.6
│   │   │   └── referencing>=0.28.0
│   │   │       ├── attrs>=22.2.0
│   │   │       └── rpds-py>=0.7.0
│   │   ├── referencing>=0.28.4
│   │   │   ├── attrs>=22.2.0
│   │   │   └── rpds-py>=0.7.0
│   │   └── rpds-py>=0.7.1
│   ├── jupyter-core
│   │   ├── platformdirs>=2.5
│   │   └── traitlets>=5.3
│   └── traitlets>=5.1
└── pyflakes

Can the heavy nb dependencies be moved to an optional extra?

I think tools like this should be kept lightweight so we have the option to install them directly in env without creating conflicts. With this tree a pipx-style installation is warranted.

[Saransh-cpp]

Oh yes, that should be done. Thanks for reporting this! General questions (for watchers and the future me) -

• should removestar throw a warning if nbconvert and nbformat is not installed?
• should this be a part of v1.6.0 or. v1.5.1?
• is this a breaking change that needs a new major version? (v2.0.0 - I don't think so)

[wimglenn]

My recommendation would be to isolate the code which requires nbconvert/nbformat into a separate submodule. That way when a user attempts to use those features without necessary deps in place you can handle the import error and print some message about installing with the [extra]. Or just let the ImportError blow up, that might be self-explanatory enough for the target audience.