prototext: Fix parsing of unknown repeated message fields

If `DiscardUnknown` was enabled, previously this syntax was rejected
with an error "unexpected token: ]":

  unknown_field: [
    {}
  ]

This is because the parser short-circuited after parsing the {}, and
didn't handle the closing square brace. Now it no longer short-circuits.

Change-Id: I6230b56766632752a5cc8822b17ed8262873d6cc
Reviewed-on: https://go-review.googlesource.com/c/protobuf/+/530616
Reviewed-by: Damien Neil <dneil@google.com>
Reviewed-by: Michael Stapelberg <stapelberg@google.com>

encoding/prototext/decode.go

@@ -739,7 +739,9 @@ func (d decoder) skipValue() error {
 			case text.ListClose:
 				return nil
 			case text.MessageOpen:
-				return d.skipMessageValue()
+				if err := d.skipMessageValue(); err != nil {
+					return err
+				}
 			default:
 				// Skip items. This will not validate whether skipped values are
 				// of the same type or not, same behavior as C++

encoding/prototext/decode_test.go

@@ -237,6 +237,11 @@ s_string: "谷歌"
 		inputMessage: &pb2.Scalars{},
 		inputText:    `unknown_field: { strings: [ [ ] ] }`,
 		wantErr:      `(line 1:29): invalid scalar value: [`,
+	}, {
+		desc:         "unknown list of message field",
+		umo:          prototext.UnmarshalOptions{DiscardUnknown: true},
+		inputMessage: &pb2.Scalars{},
+		inputText:    `unknown_field: [ { a: "b" }, { c: "d" } ]`,
 	}, {
 		desc:         "proto3 message cannot parse field number",
 		umo:          prototext.UnmarshalOptions{DiscardUnknown: true},