Completer documentation rendering by completion providers

[krassowski]

Description

Currently the renderer of the first completion provider (with highest rank) is always used. This means that if different formats of documentation are provided by different providers, the results are sub-optimal as the renderers of top provider may not know how to render other formats.

Expected behavior

Either:

• createDocumentationNode should be called on per-item instance, using the renderer of corresponding provider
• we have a good-enough default that will satisfy downstreams, i.e. allowing to render a MIME bundle (e.g. markdown with LaTeX typsetter); in this scenario dowsntreams would only override createDocumentationNode if they wanted to add some extra/dynamic details which need to live outside of the mime bundle.

Context

• JupyterLab version: 4.0.6