Follow / unfollow thread actions should bump its last activity date & time + minor general fixes

[Oaphi]

closes #1930
closes #1932
closes #1933

This PR makes sure following / unfollowing a comment thread bumps the thread's last activity date & time and thus prevents stale content from being served afterwards.

It also fixes confusing inversion of the User#anonymize method's persist_changes parameter (turns out it wasn't the best idea to rename dirty to persist_changes while tired - thankfully, the logic was sound, this change is just a matter of making it not confusing).

Additionally, it ensures attempting to make the same action multiple times does not create duplicate entries (as well as automatically cleans up any duplicate entry on unfollow).

Finally, it also makes sure we don't need to reload the whole page upon successfully unfollowing a thread (and gets us closer to removing thread_unrestrict entirely).

Note for reviewers: to be able to test this change, you need to enable caching in development (should be on by default, but it doesn't hurt to check) and set the PERFORM_CACHING env variable to true.

[codecov]

Codecov Report

❌ Patch coverage is 96.15385% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.63%. Comparing base (578ef87) to head (8a6e63b).

Files with missing lines	Patch %	Lines
app/jobs/clean_up_thread_followers_job.rb	85.71%	2 Missing ⚠️

Additional details and impacted files

Components	Coverage Δ
controllers	74.66% <100.00%> (ø)
helpers	84.91% <ø> (ø)
jobs	80.59% <85.71%> (+1.35%)	⬆️
models	90.09% <100.00%> (+0.08%)	⬆️
tasks	61.11% <ø> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[cellio]

Would it be easy to also, while we're here, make unfollow leave the in-page thread open, same as follow does? This difference was discussed in chat (pre-existing issue), but I don't know if making them consistent is easy or too much scope creep (and we should file a separate issue for it).

[cellio]

I tested (with the env var per the description) and it's working as expected. The code overall looks good to me; I have one question and would like someone else to review that part.

[trichoplax]

Having switched on caching in development as suggested (rails dev:cache from bash inside the container, followed by adding PERFORM_CACHING=true to docker/env, followed by restarting the container), I can reproduce the problem on the develop branch (both the inline thread view and the full thread page can be in a state where clicking "unfollow" appears to have no effect and even a hard refresh doesn't change it to show "follow" (or the other way around)).

On this new branch, the caching problem is fixed so if a page is not matching the server state, refreshing the page fixes it. I see that as fixing the identified problem, so I have approved this pull request, with one minor optional comment.

---

During testing I noticed a couple of pre-existing problems, so to avoid scope creep I've raised them as separate issues #1932 (which I see you already fixed while I was typing up the other one) and #1933.

[cellio]

LGTM now. Thanks for adding that route so the in-page behavior is the same for follow and unfollow.

[cellio]

Cleanup job LGTM by inspection. Re-approving pending some testing that @trichoplax is going to do. (Test code looks good; I haven't set up the bug conditions for manual testing.)