rlp: add support for optional struct fields

[fjl]

This adds support for a new struct tag "optional". Using this tag, structs used for RLP
encoding/decoding can be extended in a backwards-compatible way, by adding new fields at
the end.

[holiman]

Suggestion:

diff --git a/rlp/decode_test.go b/rlp/decode_test.go
index af4cd0acc4..c9191949fa 100644
--- a/rlp/decode_test.go
+++ b/rlp/decode_test.go
@@ -979,12 +979,17 @@ func TestInvalidOptionalField(t *testing.T) {
 	for _, test := range tests {
 		err := DecodeBytes(unhex("C20102"), test.v)
 		if err == nil {
-			t.Errorf("no error for %T", test.v)
+			t.Errorf("decode: no error for %T", test.v)
 		} else if err.Error() != test.err {
-			t.Errorf("wrong error for %T: %v", test.v, err.Error())
+			t.Errorf("decode: wrong error for %T: %v", test.v, err.Error())
+		}
+		_, err = EncodeToBytes(test.v)
+		if err == nil {
+			t.Errorf("encode: no error for %T", test.v)
+		} else if err.Error() != test.err {
+			t.Errorf("encode: wrong error for %T: %v", test.v, err.Error())
 		}
 	}
-
 }
 
 func ExampleDecode() {

[holiman]

Tested it a bit, it seems to work exactly as expected. I expected it to fail some of my tests, e..g

	type x struct {
		A uint64
		B uint `rlp:"optional"`
		C uint `rlp:"optional"`
	}

When doing EncodeToBytes(&x{0xaa, 0, 0xcc}) it handles it fine (writes all data without complaining about optional), and with EncodeToBytes(&x{0xaa, 0, 0}) it ignores the two last elements. So that's good!

[fjl]

Regarding the suggestion, I think it's not necessary because the test is specifically for the struct tags validation in typecache.go. Both encoding and decoding call into this.