[devops] CI testing is flaky

[WenheLI]

While submitting a PR, the CI auto-testing past for node 8 and failed for node 10.

This seems to be a potential issue with CI testing or test script?

[joeyklee]

@WenheLI - this is an issue that has to do with a number of things, but foremost that it takes a long time for the pre-trained models to download from google when the models are called. Usually tests pass without a problem locally, so I have a feeling this has to do with the testing environment.

@tirtawr has suggested we consider doing pseudo-testing of our functions or rather that we don't run the CI testing on PRs on development, but rather just when we do releases. Not sure yet what the best way to handle this would be.

Maybe we just pay for a better machine? Would trying Github Actions maybe help with this?

[joeyklee]

One thing I'm adding in #755 - is setting jasmine's random=false so that our tests run in sequence. This might help to address some of these funky issues!

[bomanimc]

Merging this issue into #637.