quoted printable decode swallows character when having 2 quoted values

```
>> Mail.new('Subject: "Subject:=?utf-8?B?0J/QvtC80L7Qs9C40YLQtSDQstC+0YHRgdGC0LDQvdC+0LLQ?= =?utf-8?B?uNGC0Ywg0L/QsNGA0L7Qu9GM?="')['Subject'].to_s
=> "\"Subject:Помогите восстановть пароль\"" # second word ends in ть
>> Mail.new('Subject: "Subject:=?utf-8?B?0J/QvtC80L7Qs9C40YLQtSDQstC+0YHRgdGC0LDQvdC+0LLQuNGC0Ywg0L/QsNGA0L7Qu9GM?="')['Subject'].to_s
=> "\"Subject:Помогите восстановить пароль\"" # second word ends in ить
```

Reproduce the two Mail.new examples in a Ruby console and compare the decoded Subject values. Trace the quoted-printable decoding path used by the mail parser; done means both examples preserve the complete final characters, including the difference between ть and ить.

Written by the indexing model from the issue text.