Optimize load/store with an iadd_imm operand.

Fold the immediate into the load/store offset when possible.

cranelift/filetests/postopt/fold_offset_into_address.clif

@@ -0,0 +1,18 @@
+test postopt
+target x86_64
+
+; Fold the immediate of an iadd_imm into an address offset.
+
+function u0:0(i64 vmctx) -> i64 {
+ebb0(v0: i64):
+                     v1 = iadd_imm.i64 v0, 16
+[RexOp1ldDisp8#808b] v2 = load.i64 notrap aligned v1
+[Op1ret#c3]          return v2
+}
+
+; sameln: function u0:0(i64 vmctx) -> i64 fast {
+; nextln: ebb0(v0: i64):
+; nextln:                                     v1 = iadd_imm v0, 16
+; nextln: [RexOp1ldDisp8#808b]                v2 = load.i64 notrap aligned v0+16
+; nextln: [Op1ret#c3]                         return v2
+; nextln: }