plotly core (a community vote for a simple wrapper to plotly.js)

Hello, long time user of plotly.R here. I thought this issue could start the discussion to split up the package into a core and a full ggplotly supported package.

As many have noted, I also started using plotly from the ggplot2 interface of ggplotly,
but as times are changing, and I am getting to understand the drawbacks of wrapping things through ggplot more,
we are moving more and more away from it.

In our RiboCrypt app, we have for optimizations reasons removed all ggplotly calls and transformed to native plot_ly() calls. We experiment with optimizations like pre-building plotly template objects, to speed up "time to user see something useful".
And since this is an app for complicated genomics, R/bioconductor is still a far better fit than python.

A problem has been the last few year that new plotly.R versions break stuff, so I have to keep version set to a version I know works. Problem is that new version needs to pass not just plotly.js tests, but also ggplotly tests for regression, so it is currently much more complicated than it needs to be.

I vote for a clean split of the package, for existing popular R package doing this, see for example the fst package, which has fst and fstcore

plotlycore: Simple wrapper to plotly.js
plotly: Uses plotlycore + ggplotly interface.

This way I think it is more safe especially for people using it for actual production apps and not just scripting.

I thought we could discuss here what people have noticed lately, what the relevant goals of current maintainers are, what resources exist to do this etc.

Beginne damit, die vorgeschlagene Aufteilung zwischen einem einfachen Wrapper für plotly.js und dem ggplotly-unterstützten Paket zu lesen, und vergleiche anschließend das in der Issue verlinkte fst/fstcore-Beispiel. Prüfe die aktuelle Struktur des plotly.R-Pakets sowie die Verantwortlichkeiten von plot_ly(), ggplotly und regression-test. Als erledigt gilt die Aufgabe, wenn sich die Maintainer auf konkrete Paketgrenzen und Implementierungsziele geeinigt haben.