[llvm] [LangRef] Specify that syncscopes can affect the monotonic modification order (PR #189017)

[Fabian Ritter via llvm-commits]

ritter-x2a wrote:

> Proposed a simpler wording that has fewer phrases to defend.

​Thank you for the suggestions, I think those work for me. The point I wanted to bring across is that operations with mismatching syncscopes still can (but don't have to) be in those relations, since the orders are transitive.
For example, if operations A, B, and C are all seqcst and only (A,B) and (B,C) (but not (A,C)) have inclusive scope, then transitivity of the seqcst order still means that (A->C) must be related if (A->B) and (B->C) are in the seqcst order.
I think your suggestions capture that sufficiently (and make the text a lot more readable), so I've updated the PR with them.

This wording is also related to the [discussion about the need to specify what guarantees the default syncscope provides in a previous PR](https://github.com/llvm/llvm-project/pull/189015/changes#r3069421358).

https://github.com/llvm/llvm-project/pull/189017