Fix flaky redo document capture feature test 100% for sure, no chance of failure

[matthinz]

#8576 added a new feature test to check that, when using the "redo document capture" action, users are not able to access earlier steps of the IdV process (doc capture, ssn, etc.) after submitting the "Verify info" screen. We added this new test in among some existing redo document capture feature tests.

After adding the new test, we noticed it was flaky in CI. We temporarily disabled the new test in #8607, but the flakiness remained with other tests in the same file. Our current hypothesis is that some of the refactoring we did to allow adding the new test in amongst the existing tests has led to the flakiness.

This PR:

• Reverts the redo document capture feature tests back to their state pre-LG-9996: Don't allow redoing doc capture after verify info #8576 (with changes added in the meantime in LG-8830 redo_document_capture logging #8581 preserved)
• Adds the test from LG-9996: Don't allow redoing doc capture after verify info #8576 back, but in a new, independent context to minimize interactions with other tests in the file.

[matthinz]

(I'm going to run CI several times on this branch before marking ready to merge.)

[aduth]

Looking at an example failure, I'd suspect a reason it may fail is because the submission of document capture involves an asynchronous submission we're not waiting to complete.

It seems like if we substitute the attach and submit behaviors with the attach_and_submit_images helper, it may help, since the helper explicitly waits for the load screen to disappear?

[matthinz]

It looks like we're using attach_and_submit_images here. The click_idv_continue is to move past the "We couldn’t read the barcode on your ID" screen to get to the SSN entry screen. So we are waiting, but possibly just not waiting long enough in CI sometimes? Though 10 seconds seems like plenty of time to wait. I'll try lowering that locally and see if I can get a failure.

[aduth]

Oh, d'oh, I totally misread the test setup. That's what I get for commenting so early in the morning 🤦 You're right about it already using attach_and_submit_images.

That is strange, since it doesn't seem like there's anything asynchronous in clicking "Continue" from the barcode warning page that would delay the page transition (source).

[matthinz]

Gonna close this PR--I'm working on this in the background, and we have a ticket now (LG-10192), but this PR is just noise ATM.