Sharing PR Modeling

[danieljurek]

After fixing the acute issue in LintDiff and discovering and fixing a couple more failure scenarios, let's have a closer look at how LintDiff models PRs and whether we can use that capability in other places like Breaking Changes.

• Parameterize tests or come up with a Domain-Specific Language?~ DONE
• Is there some overlap with Breaking Changes?
• There's an opportunity for a "PRModel" which can compare before/after in 2 different SpecModel objects and give useful results
• Maybe factor the tests so that the scenarios can be easily extracted and used in other tests

Originally posted by @danieljurek in #38698 (comment)