Address CI issues related to GDrive download quota

[NivekT]

🚀 The feature

Mitigate false positive test failures related to Google Drive. Some possible solutions are:

1. Move the files to a location that is still on GDrive but has a higher daily download quota (or no quota at all)
   • We can also use different files that are already in such GDrive location.
2. Make a request to third-party to life the quota for particular files

Motivation, pitch

Within the tests that have the decorator @slow (i.e. tests triggered by Github label "ciflow/slow"), a few of them connect to Google Drive to download files in order to test our DataPipes' functionalities. These tests occasionally fail because of issues related to download quota and not because of a bug in our implementation.

Alternatives

Change our tests/codes to ignore these cases, but that may just end up not testing the functionalities that we want to test.

Examples

CI failures from a PR

[ejguan]

We may be able to use the API from google https://developers.google.com/drive/api/v3/about-sdk to see if this can be bypassed.

[parmeet]

We have been long struggling with this issue on torchtext. As a middle-ground, we are doing daily caching on CircleCI that caches all the datasets in on job, and that is used by all other jobs/PR. But we are moving away from this in favor of mocked tests for our datasets pytorch/text#1493