As others have said, they are pretty similar at this point, which is not by accident as the Juno team explicitly aimed for feature parity (see the issue label here), although there are also still open issues where things aren’t on par with Juno (see issues here, although not all of them actually describe a working Juno feature).
Juno is currently in maintenance mode, i.e. it receives bug fixes but no feature development, which means that the VSCode extension now has some nice stuff that is unlikely to ever be available in Juno (fwiw I don’t seem to use much of it, but a good example of something recent and really cool is the native Jupyter notebook support).
One thing to note is that while on the surface things look pretty similar, there’s a pretty big difference in underlying design: VSCode is designed as a tool that statically extracts as much information as possible from the code you’ve written to give you autocompletes, linting, automatic insertion of using statements etc., but it doesn’t have any dynamic insight into your running session. Juno on the other hand is tightly integrated with your current running session, which has its issues (it has to load packages into your session at startup, which sometimes has unintended consequences) but also means that e.g. you can autocomplete column names of a DataFrame, which are not inferrable from just the code you’ve written.
If you’re happy with Juno, there’s no reason to switch, just swap out the Julia version in the Atom settings and you’re done. If you’re a grad student looking to tinker with your tooling to procrastinate, I’d say switching is worthwhile as VSCode is probably more future-proof and might have (or in future get) some cool feature which won’t make it into Juno.