[llvm] [msan] Add experimental '-msan-or-shadow-for-strict-instructions' flag to pessimize output (PR #128036)

[Florian Mayer via llvm-commits]

fmayer wrote:

> for non-recover mode: the shadow will now be properly propagated instead of being reset to clean after each unknown instruction.

I don't understand how we can "properly" propagate for unknown instructions.

https://github.com/llvm/llvm-project/pull/128036