Add kernel independent logging

[gclen]

This moves jupyter/notebook#2252 to jupyter_server. I think it is useful to have the ability to log which code has been executed (for auditability reasons).

The method of specifying a python logger is clunky and I'm looking for suggestions on ways to improve that (specifying the logger in a traitlets config file?).

Pinging @Zsailer

[Zsailer]

Thanks, @gclen! Pinging @kevin-bates for his thoughts, too.

[kevin-bates]

Thanks for starting this @gclen.

I like the idea and think this could be generally useful - especially in larger installations. It feels like a great fit as an event source for the Jupyter Telemetry project and makes sense to leverage for this.

I would imagine we'd include more data about the kernel "session" (user, kernel-id, kernel-name, provider, etc.) that could then be used to differentiate activity, either via cross-referencing (via kernel-id) or directly.

[Zsailer]

Sorry—I didn't have time to look at this closely yesterday. I thought this was directed at developers for debugging purposes, not at operators/administrators auditing the server.

I agree with Kevin. This should be handled by telemetry instead.

I'll draft a PR that adds telemetry to the server this afternoon.

[Zsailer]

See #233 for a draft of telemetry/event logging in Jupyter server.

[Zsailer]

I'm going to close this PR in favor of #233.

That said, @gclen it would be great to work with you on developing an event schema for the kernels API. As someone who is interested in logging/auditing kernels, you have super valuable insight there.

[Zsailer]

Checkout #238 for initial discussion on drafting a kernels schema.