[llvm] [GlobalISel] Use G_ANYEXT instead of G_ZEXT for shift amount widening (PR #223136)

[Matt Arsenault via llvm-commits]

================
@@ -14,7 +14,9 @@ define <8 x i16> @dupsext_v8i8_v8i16(i8 %src, <8 x i8> %b) {
 ; CHECK-GI-LABEL: dupsext_v8i8_v8i16:
 ; CHECK-GI:       // %bb.0: // %entry
 ; CHECK-GI-NEXT:    lsl w8, w0, #8
-; CHECK-GI-NEXT:    sbfx w8, w8, #8, #8
+; CHECK-GI-NEXT:    mov w9, #8 // =0x8
+; CHECK-GI-NEXT:    sxth w8, w8
----------------
arsenm wrote:

This increased the instruction count 

https://github.com/llvm/llvm-project/pull/223136