Draft: Suggestion: Improve automatic MR feedback

DangerBot feedback might be visually challenging to grasp both to contributor and reviewer. This is a draft suggestion to improve the DangerBot feedback form.

Currently we put into the comment:

• Automated warnings in a form of a table
• Review suggestions and related warnings (DB review, Docs review, reviewer roulette suggestions)
• Code warnings (like multiversion compatibility concern)

Suggestions:

• Separate review suggestions from MR merge warnings (maybe into two comments)
• Require a thread resolution on the warnings
• Reprioritise notifications (master failures are not blocking warnings)
• Format the comment in a consistent manner (not table + paragraphs + roulette)
• Reduce the noisiness by hiding extra details

Also, can we rename the 🚫 Danger to something softer? Feels like we use "Danger" when it's not needed. Which can influence how we perceive the word.