x64: fix register allocation panic due to load-coalesced value (#3954)

Fuzz testing identified a lowering case for CLIF's `icmp` in which the
double use of a loaded operand resulted in a register allocation error.
This change manually adds `put_in_xmm` to avoid load-coalescing these
values and includes a CLIF filetest to trigger this issue. Closes #3951.

I opened #3953 to discuss a way in which this kind of mistake (i.e.,
forgetting to add `put_in_*` in certain situations) could be avoided.

cranelift/filetests/filetests/isa/x64/simd-issue-3951.clif

@@ -0,0 +1,19 @@
+test compile
+set enable_simd
+target x86_64 skylake
+
+;; Compile a CLIF version of the register allocation issue identified in
+;; https://github.com/bytecodealliance/wasmtime/issues/3951.
+function %check_issue_3951(i64 vmctx) -> i8x16 fast {
+    gv0 = vmctx
+  block0(v0: i64):
+    v2 = global_value.i64 gv0
+    v3 = load.i8x16 notrap aligned v2+8
+    v4 = global_value.i64 gv0
+    v5 = load.i8x16 notrap aligned v4+8
+    v6 = icmp ugt v3, v5
+    v7 = raw_bitcast.i8x16 v6
+    jump block1(v7)
+  block1(v1: i8x16):
+    return v1
+}