Migrate Paparazzi to Compose Preview Screenshot testing

[isaacnguyen0809]

Pull request (PR) checklist

Please check if your pull request fulfills the following requirements:

• [x] The PR is submitted to the main branch.
• [x] I've read the Contribution Guidelines and my PR doesn't break the rules.
• [x] I've read and understand the Developer Guidelines.
• [x] I confirm that I've run the code locally and everything works as expected.
• [x] My PR includes only the necessary changes to fix the issue (i.e., no unnecessary files or lines of code are changed).
• [ ] 🎬 I've attached a screen recording of using the new code to the next paragraph (if applicable).

Description

• Let's check the integration first and if it's fine, i will start write the docs for it

What's changed?

• Migrate Paparzzi testing to Compose preveview screenshot testing
• Remove Paparazzi integration
• Add image difference threshold

Risk factors

...

Does this PR close any GitHub issues?

• Closes #3243

Troubleshooting CI failures ❌

Pull request checks failing? Read our CI Troubleshooting guide.

[ILIYANGERMANOV]

Nice! Google have added threshold support. A few notes:

• currently 100% threshold makes the tests useless (if I understood it correctly)
• We neee to update the docs before merging, otherwise contributors will start pinging me

[isaacnguyen0809]

@ILIYANGERMANOV - please ignore this comment for now, i need to check it again

According to what I have tried, an image difference threshold of 6% can pass all cases. The case that raises the threshold to 6% is due to the Disclaimer Screen. I will attach the test results in the file below, with a 3% threshold, it passes only 50%. Excluding this screen, I find that the remaining test cases can pass with a threshold of 1.3%.

Therefore, i think we have two options:

1. We accept a 6% threshold (if you’re comfortable with this)
2. Because the Disclaimer Screen is just a static screen showing text without any complex UI, we can skip the screenshot test for this screen and use the 1.3% threshold for the rest. Wdyt?

BTW, I have updated the docs for this part. Could you check if anything else is needed?

[ILIYANGERMANOV]

Nice work @isaacnguyen0809! I think 2) is a good option. The best choice would be to set 1.3% for all tests and 6% exception for the Disclaimer (e.g. support threshold customization for different tests). Or we can find what makes the Disclaimer screen flaky.

Anyway, I'm happy with skipping its tests if my suggestions are too hard to impl

[ILIYANGERMANOV]

We can re-open it when done