Missed Simplification of trunc (lshr (((shl %1, 1) | 1) + %2), 1) to i8 into trunc (%1 + lshr (%2 + 1), 1) to i8

The following reduced IR is derived from https://github.com/PixarAnimationStudios/OpenUSD/blob/9d7fa210/pxr/imaging/plugin/hioAvif/aom_dsp/loopfilter.c#L378
Godbolt: https://godbolt.org/z/xaTaGMcEM
alive2 proof: https://alive2.llvm.org/ce/z/S2Gfd6

```llvm
define void @filter14(ptr %0, i32 %1, i32 %2) {
%reass.add = shl i32 %1, 1
%4 = or i32 %reass.add, 1
%5 = add i32 %4, %2
%6 = lshr i32 %5, 1
%7 = trunc i32 %6 to i8
store i8 %7, ptr %0, align 1
ret void
}
```

expected:
```llvm
define void @tgt(ptr %0, i32 %1, i32 %2) {
%4 = add i32 %2, 1
%5 = lshr i32 %4, 1
%6 = add i32 %5, %1
%7 = trunc i32 %6 to i8
store i8 %7, ptr %0, align 1
ret void
}
```

Start by reproducing the reduced LLVM IR from the issue with the linked Godbolt example and compare the result with the Alive2 proof. Trace which LLVM simplification or instruction-combining entry point handles the shift, add, or truncation sequence, and confirm completion when the generated IR matches the expected form.

Written by the indexing model from the issue text.