Safe handling of plugin installations using as source PyPI for bundle/conda based napari installations

[dalthviz]

Currently the only measure in this regard is setting as default source conda in case the installation is conda based/from bundle and a conda package existis for the plugin. Mixing pip and conda can be harmful for installation stability in general so it would be nice to have a way to ensure that mixing them is handled in a way that potencial conflicts are prevented/detected (for example installing all the dependencies required by the plugin from conda and only installing the plugin itself from PyPI).

Although in the future proper handling could be added, as things are, some initial ideas to partially mitigate potencial errors are:

• Not show the PyPI option from the source combobox when running from a bundle installation (basically prevent installing plugins from PyPI if using a bundle based installation).
• Add a warning if the PyPI source option is selected when running from a bundle installation: it could be a dialog to confirm the installation for example and it could also potentially have a link to some docs page explaining why mixing pip + conda could be harmful and some other useful related resources like how to make a plugin available from conda-forge (maybe https://napari.org/stable/plugins/testing_and_publishing/deploy.html#plugin-deploy with some updates).

For more details you can see the discussion at https://napari.zulipchat.com/#narrow/channel/309872-plugins/topic/Can't.20find.20my.20plugin.20in.20napari.20bundled.20app.2E/near/480006024

[psobolewskiPhD]

@jaimergp linked this in the meeting yesterday:
https://conda-incubator.github.io/conda-pypi/
New to me, but I think this nicely adresses most of the issues on the bundle side and in the case where napari was installed via conda.
Frankly, I'd be pretty OK with using a still-experimental conda-pypi in the bundle as a proof of concept.

[jaimergp]

With this #111 having been merged I think we are in a good spot, but let's keep this open for experimentation with conda-pypi and others.