[clang] [Clang] Ignore assumptions with side effects at compile time (PR #85534)

[via cfe-commits]

Sirraide wrote:

> This is not really an NFC change so you should have waited for an approval.

Yeah, that’s something I wasn’t sure about in this particular case: The code that was causing the problem was added and merged in by me earlier this week, so I figured it was fine (and this seemed like an ‘obvious’ fix seeing as this is what we’re also checking for in codegen), but I can wait for approval next time if a situation like this comes up again.

https://github.com/llvm/llvm-project/pull/85534