Connect to jupyter kernels

[shibbas]

This feature enables connecting to Jupyter kernels via the #!connect jupyter command and interact with them to perform the following:

• Handle commands: SubmitCode, RequestCompletions, RequestSignatureHelp, RequestHoverText
• Add support for Variable sharing, including dataframes, using Jupyter Comms for python and R. This setup can be used to enable variable sharing for other Jupyter kernels that support comms as well by providing just the kernel side implementation to handle the SendValue, RequestValue and RequestValueInfos command.
• Enable connecting to locally installed kernels directly over ZMQ using command:

#!connect jupyter --kernel-name python --kernel-spec python3

This currently does not support selecting environment yet. vscode will need to start in the environment with Jupyter installed.

• Enable connecting to remote kernels over HTTP and websocket using command:

#!connect jupyter --url <url_for_jupyter> --token <jupyter_server_token> --kernel-name python --kernel-spec python3

Known issues:

• stdin via input() is disabled. Use the input prompt in polyglot and share the value instead: https://github.com/dotnet/interactive/blob/main/docs/input-prompts.md
• For ZMQ kernels, execution in kernel is not stopped when the cell run is canceled.

[shibbas]

We might want to revisit this design to support the cold start scenario.