Add support for arbitrary Jupyter Kernels

There are a whole lot of Jupyter kernels. Like, hundreds of them. Currently -- despite the fact that Positron's own R and Python subsystems run on Jupyter kernels -- you can't hook up a random Jupyter kernel up to Positron's front end. You need to, at a minimum, write a little extension that advertises your kernel as a language runtime, wraps the Jupyter Adapter to start/manage your kernel session, etc.

It would be interesting to make it possible to hook up a Jupyter kernel to Positron in a tool-free way, such that an end user could point Positron at their installed Jupyter kernel and have it start working immediately.

Of course, these sessions wouldn't get the Full Positron Experience: you'd only have static plots, no Variables pane or Data Explorer, etc. But just being able to load these kernels as REPLs/code execution engines would be very valuable for long-tail languages, and/or those for which no formal language pack has been defined.

The issue names Positron's front end, its Jupyter Adapter, and the existing R/Python subsystems, but no files or tests. Start by mapping how those subsystems advertise runtimes and manage Jupyter sessions; done means an installed arbitrary Jupyter kernel can be selected and used for REPLs or code execution without an extension, with the stated feature limitations.