Add custom legalization for floating point constants.

Use the simplest expansion which materializes the bits of the floating
point constant as an integer and then bit-casts to the floating point
type. In the future, we may want to use constant pools instead. Either
way, we need custom legalization.

Also add a legalize_monomorphic() function to the Python targetISA class
which permits the configuration of a default legalization action for
monomorphic instructions, just like legalize_type() does for polymorphic
instructions.

cranelift/filetests/isa/intel/legalize-custom.cton

@@ -52,3 +52,19 @@ ebb0(v1: i32):
     ; check: $new:
     ; nextln: return
 }
+
+function %f32const() -> f32 {
+ebb0:
+    v1 = f32const 0x1.0p1
+    ; check: $(tmp=$V) = iconst.i32
+    ; check: $v1 = bitcast.f32 $tmp
+    return v1
+}
+
+function %f64const() -> f64 {
+ebb0:
+    v1 = f64const 0x1.0p1
+    ; check: $(tmp=$V) = iconst.i64
+    ; check: $v1 = bitcast.f64 $tmp
+    return v1
+}