fix handling of RFC2047 words containing invalid bytes after decoding

Yet another bad case somehow not covered by all the tests. If the
encoding is recognised, and the word is successfully decoded, but the
resulting string contains an invalid byte sequence for its declared
encoding, Sup would crash with ArgumentError.

Only use the decoded word if the end result has a valid encoding.

Author: danc86

lib/sup/rfc2047.rb

@@ -62,10 +62,13 @@ def Rfc2047.decode_to(target, from)
       end
 
       begin
-        text.force_encoding(charset).encode(target)
+        text.force_encoding charset
+        text.encode! target
       rescue ArgumentError, EncodingError
-        word
+        next word
       end
+      next word unless text.valid_encoding?
+      text
     end
   end
 end

test/fixtures/rfc2047-header-encoding.eml

@@ -1,4 +1,4 @@
-From: test@example.invalid
+From: =?utf-8?q?YouTube-tj=E4nst?= <service@youtube.com>
 To: test@example.invalid
 Date: Sun, 19 Jul 2020 17:03:56 +1000
 Subject:

test/test_message.rb

@@ -253,6 +253,7 @@ def test_rfc2047_header_encoding
                  "bad: =?UTF16?q?badcharsetname?==?US-ASCII?b?/w?=" +
                  "=?UTF-7?Q?=41=6D=65=72=69=63=61=E2=80=99=73?=",
                  sup_message.subj)
+    assert_equal "=?utf-8?q?YouTube-tj=E4nst?=", sup_message.from.name
   end
 
   def test_nonascii_header