[clang] [analyzer] Suppress out of bounds reports after weak loop assumptions (PR #109804)

[Gábor Horváth via cfe-commits]

Xazax-hun wrote:

Some random, unorganized questions:
* Any reasons we cannot make block counts available to bug reporters and get a best of both worlds (relatively simple implementation and get the separation of concerns)?
* Does any of the approaches have trouble dealing with short-circuiting operators on the loop condition?
* Would it make sense to have a combined approach? E.g., the engine adding `ProgramPointTag` in the engine where weak assumptions happened, and using a bug reporter visitor to look for those tags. I wonder if that approach would make reading exploded graphs easier when investigating false positives. 

https://github.com/llvm/llvm-project/pull/109804