cmd/compile: simplify value coding for unified IR

In indexed export, values are always exported along with their type
and are encoded in a type-sensitive manner, because this matches how
cmd/compile handled constants internally.

However, go/types intentionally differs from this, decoupling type
from value representation. As unified IR strives to be more
go/types-centric, it makes sense to embrace this and make values a
more first-class encoding.

Change-Id: If21d849c4f610358bd776d5665469d180bcd5f6e
Reviewed-on: https://go-review.googlesource.com/c/go/+/348014
Trust: Matthew Dempsky <mdempsky@google.com>
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Cuong Manh Le <cuong.manhle.vn@gmail.com>

src/cmd/compile/internal/noder/writer.go

@@ -542,7 +542,8 @@ func (w *writer) doObj(obj types2.Object) codeObj {
 
 	case *types2.Const:
 		w.pos(obj)
-		w.value(obj.Type(), obj.Val())
+		w.typ(obj.Type())
+		w.value(obj.Val())
 		return objConst
 
 	case *types2.Func:
@@ -598,12 +599,6 @@ func (w *writer) typExpr(expr syntax.Expr) {
 	w.typ(tv.Type)
 }
 
-func (w *writer) value(typ types2.Type, val constant.Value) {
-	w.sync(syncValue)
-	w.typ(typ)
-	w.rawValue(val)
-}
-
 // objDict writes the dictionary needed for reading the given object.
 func (w *writer) objDict(obj types2.Object, dict *writerDict) {
 	// TODO(mdempsky): Split objDict into multiple entries? reader.go
@@ -1199,7 +1194,8 @@ func (w *writer) expr(expr syntax.Expr) {
 
 			w.code(exprConst)
 			w.pos(pos)
-			w.value(tv.Type, tv.Value)
+			w.typ(tv.Type)
+			w.value(tv.Value)
 
 			// TODO(mdempsky): These details are only important for backend
 			// diagnostics. Explore writing them out separately.