[clang] Thread Safety Analysis: Decode terminator branches directly into capability resolutions [NFC] (PR #220634)

[Jameson Nash via cfe-commits]

vtjnash wrote:

> I don't think this is an NFC change. It changes performance properties of the implementation, and introduces new concepts.

I had that same argument with Claude. It was convinced I should call this NFC because it argued that those are implementation details, not functional details, which I eventually thought was a good enough point to admit keeping the tag. I'm definitely willing to remove it if that seems clearer.

https://github.com/llvm/llvm-project/pull/220634