Restore behavior of the "raises-exception" cell tag

[joelostblom]

This restores the intended behavior of the raises-exception cell tag. Details can be found in #7015.

Close #7015

@vidartf Let me know if there are more changes required.

[jupyterlab-dev-mode]

Thanks for making a pull request to JupyterLab!

To try out this branch on binder, follow this link:

[vidartf]

All tests pass, but I won't be able to give this a spin this week (the binder doesn't have working kernels). For reviewer (non collaborators can also review): A simple test is to make two cells and run with "Run all":

5 / 0

and

5 + 5

The second cell should run with output, and the first cell get traceback for the zero division.

[jasongrout]

@vidartf, @joelostblom - are we sure we want to include all the cell metadata in an execute request? It seems there would be a lot of superfluous stuff in there. Perhaps we can just include the things we expect to be meaningful to an execute request?

[jasongrout]

I see that the output execution method is looking into this metadata to find this tag. But then all metadata is being passed on to the execution. This seems loose, passing on all sorts of irrelevant things into the actual execution message.

[jasongrout]

@blink1073 and I discussed this. The difficult thing here is that we want to tell some things to the output execute function, and some things to the kernel, and right now we are combining those things in the output execute arguments.

So let's merge this for now as a bugfix that we can backport to 1.0.x. Then in a separate PR we can introduce a new argument option to the output execute function for this information, and then we can explicitly talk to the output execute function vs the kernel message.

[blink1073]

@meeseeksdev backport to 1.0.x

[joelostblom]

Thanks for merging so quickly!