Lazy load strategy for plotly bundle

[nrbgt]

While having the plotly bundle be deduplicated between this and other extensions (#1891) would help some of the issues users are seeing, the fact still remains that, minified or unminified, the plotly bundle is enormous: even minified, it's 4mb size is close to the entire minified lab bundle (5.5mb, last i checked).

Not familiar with this code base and whichever other ones use the plotly bundle, the example for another heavy widget #1891 (comment) would at least get the bundle out of the critical path. It is likely worth investigating if there is a way to do it at a finer-grained level (e.g. just plotly itself) but then the additional points of asynchrony might need to be found.

As suggested in #1891 (comment), lazy loading the bundle would provide several benefits to end users:

• less up-front time waiting for the "moons" animation for all users
• less CPU used during a local webpack rebuild for those that aren't using hosted/pre-built lab applications

The only downside is there would be a (potentially noticeable on slower connections) delay when the first widget is requested.

[pragyagarg642]

I have actually done this in my local. I used dynamic imports to lazy load the plotly bundle in jupyterlab-plotly extension. It seems to be working fine on my local machine. Shall I submit a PR?

[nicolaskruchten]

This is (finally) closed as part of #3142