LTI replace results on every score

[kiragrammel]

Closes #1648

[codecov]

Codecov Report

Attention: 23 lines in your changes are missing coverage. Please review.

Comparison is base (be30666) 66.74% compared to head (3edbb60) 69.77%.

❗ Current head 3edbb60 differs from pull request most recent head 2a0d632. Consider uploading reports for the commit 2a0d632 to get more accurate results

Files	Patch %	Lines
.../controllers/user_exercise_feedbacks_controller.rb	15.00%	17 Missing ⚠️
app/controllers/community_solutions_controller.rb	0.00%	2 Missing ⚠️
app/controllers/exercises_controller.rb	0.00%	1 Missing ⚠️
app/controllers/submissions_controller.rb	93.75%	1 Missing ⚠️
app/models/user_exercise_feedback.rb	50.00%	1 Missing ⚠️
app/policies/user_exercise_feedback_policy.rb	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1698      +/-   ##
==========================================
+ Coverage   66.74%   69.77%   +3.03%     
==========================================
  Files         195      196       +1     
  Lines        6182     6168      -14     
==========================================
+ Hits         4126     4304     +178     
+ Misses       2056     1864     -192     

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

[MrSerth]

Awesome, nicely done and many thanks for your perseverance! 👏 🥳 🎉 🥂 👯

I am very that you took the challenge to work on this crucial part of the application while digging deep into everything that's LTI related and covering your changes with so many, extensive tests. As you see, almost each line touched is covered, and a magnitude more lines are newly covered by tests even without any change.

Furthermore, I just finished my review of all code parts and am really satisfied with the result. For a final test, I've deployed the changes to Staging and CodeMoon, allowing us to test-drive the changes in a real-world context.

When doing so myself, I discovered another error with the user exercise feedbacks, not caused by our recent change. Besides that, I was surprised about the 10% rule of the feedbacks, because it was much more frequent in my tests. Regardless of that, the basic workflow was as expected. Nevertheless, it probably makes sense to have a final check 🤓.