Real Jupyter functional tests on Windows take far too long

[IanMatthewHuff]

I've had to bump up the timeout way up for the nightly flake suites. And it's just windows causing this issue. My quick guess is that it's the "locate the right jupyter to launch" code, as I think that multiple SxS versions of Python are installed on the test boxes.

[karthiknadig]

Is there any thing that extension team should investigate here? This does not look like something we can address.

[DonJayamanne]

Oops that's was accidental, adding DS label back

[rchiodo]

Following the directions here:
https://stackoverflow.com/questions/42997898/how-to-enable-source-map-support-for-node-js-prof-process

and by making a new test entry like this in the package.json:

        "test:functional:perf": "node --inspect-brk ./node_modules/mocha/bin/_mocha --require source-map-support/register --opts ./build/.mocha.functional.perf.opts",

I can profile the tests.

[rchiodo]

Additionally using this is also possible, just not as nice of an output:
https://nodejs.org/en/docs/guides/simple-profiling/

[rchiodo]

Looks like the majority of the time is enumerating interpreters by asking the registry. We do this on every test.

Hotspots:

• getInterpretersFromRegistry
• getInterpreterDetails

[IanMatthewHuff]

Validated via the most recent test results timing. Now just 2x linux mac and only 1 hour.