[llvm] [ARM][Thumb1] Account for tail duplication in fn size estimation (PR #214228)

[Simon Tatham via llvm-commits]

statham-arm wrote:

That is the kind of reason why I [originally suggested](https://discourse.llvm.org/t/rfc-arm-fixing-function-size-estimation/90626) abandoning this "estimation in advance" strategy and trying something else! I admit I hadn't quite expected _that_ many in-between passes to worry about, though.

https://github.com/llvm/llvm-project/pull/214228