Add downstream tests against `jupyter_server` as a part of release workflow

Problem

jupyter-server depends on tornado and underpins JupyterLab, Notebook and the Jupyter server extensions. Because a Python environment can hold only one version of a package, jupyter-server cannot add an upper bound on tornado as any other package in the environment needing the newer tornado would make the whole environment unresolvable. So jupyter-server, like most Python projects, pins lower bounds only, and every tornado release reaches every Jupyter user as soon as it is published which leaves Jupyter exposed to any breaking release. jupyter-server does test against its dependencies daily but that is reactive: a release like 6.5.9, which had no pre-release, is only seen once it is already out.

Proposed solution

Taking into account that, based on PyPI numbers, jupyter-server is one of the major tornado consumers, would it make sense to have a job as part of the release process that tests jupyter_server with the version about to be released?