feat(tables): update samples for making prediction with exported model

[helinwang]

Exported models runs inside a docker image in localhost. This commit
adds code sample for making prediction with exported model with
Python. It will be used in
https://cloud.google.com/automl-tables/docs/model-export#automl-tables-example-drest

[telpirion]

It looks like this sample is missing tests. Would you please add them?

[busunkim96]

Please add a test to https://github.com/GoogleCloudPlatform/python-docs-samples/blob/4af4f19da3bc8cc33764bb7bc26ca9a535c81aae/tables/automl/predict_test.py

[sirtorry]

this is going to be part of a tutorial where the reader will be asked to stand up a local server before running this snippet. how should we test it? should we test it?

[helinwang]

Maybe we can test the correctness of the request JSON example.

[busunkim96]

Passing along testing suggestions from @tmatsuo and @kurtisvg in chat:

kurtisvg

My first attempt would be just to mock the response to the request being sent to the local server

My second suggestion (and perhaps a little more e2e, but also potentially more flaky) would be to create a pytest fixture > that runs the server... I think we have some other tests that already use docker in docker but I'm not sure.

tmatsuo

docker in docker should work with a caveat. If you mount files when running the image, the path should be under /tmp. For example, this docker run -v /somewhere/mydir:/mydir hello doesn't work but docker run -v /tmp/mydir:/mydir hello should work

[busunkim96]

Closing since this PR has gone stale. These samples have been moved to the python-automl repository.