Or, adding in a devcontainer.json setup?

Sure, I think that could be nice (although it may still be painful to develop with mlpack on Windows!).

Originally posted by @rcurtin in #3603 (comment)

The goal of providing a devcontainer is to allow create a full-featured development environment within VS Code that:

1. Can be used locally if Docker and VS Code are installed;
2. May be used with GitHub's Codespaces (Docs) to instantly spin up a valid development environment.

For instance, if you click:

Then, a codespace will instantiate with a pre-configured environment for RStudio Server. For more details, see: https://github.com/coatless-devcontainer/rstudio-server

With this being said, the devcontainer spec provides many different configuration options. So, we can quickly customize as needed.

The main concern would be what should be the base image being used. I think we can re-use one of the docker images published for Jenkins.

At a later time, we might also want to consider different devcontainer setups that load the necessary R/Go/Python/Julia components pre-installed.