Updated encoded-word generation. Previously a minimal value was encoded, so it was possible to have multiple encoded words in a string separated by non encoded-words. This was an issue with some webmail clients that stripped out the non-encoded parts between encoded-words so the updated method uses wide match by encoding from the first word with unicode characters to the last word. "a =?b?= c =?d?= e" -> "a =?bcd?= e"