[wasm-simd][ia32] AVX I16x8SConvertI32x4 can have non-register input

In instruction selector for this opcode, when AVX is supported, we don't
require the second operand to be a register, so it can be an operand.

Bug: v8:1155497,v8:9198
Change-Id: I609deb69b826e686dddc237dcf2ab6129424e761
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2575787Reviewed-by: Bill Budge <bbudge@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#71650}

src/compiler/backend/ia32/code-generator-ia32.cc

@@ -2933,7 +2933,7 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
     case kAVXI16x8SConvertI32x4: {
       CpuFeatureScope avx_scope(tasm(), AVX);
       __ vpackssdw(i.OutputSimd128Register(), i.InputSimd128Register(0),
-                   i.InputSimd128Register(1));
+                   i.InputOperand(1));
       break;
     }
     case kSSEI16x8Add: {