include_headers has to be a list of bytes on py3

[Daniel Hahler]

I've noticed the following test failure with django-ses's tests, when dkimpy is used:

```
Traceback (most recent call last):
  File "…/Vcs/django-ses/tests/test_backend.py", line 177, in test_dkim_mail
    send_mail('subject', 'body', '<email address hidden>', ['<email address hidden>'])
  File "…/Vcs/django-ses/.tox/py36-django111-dkim/lib/python3.6/site-packages/django/core/mail/__init__.py", line 62, in send_mail
    return mail.send()
  File "…/Vcs/django-ses/.tox/py36-django111-dkim/lib/python3.6/site-packages/django/core/mail/message.py", line 348, in send
    return self.get_connection(fail_silently).send_messages([self])
  File "…/Vcs/django-ses/django_ses/__init__.py", line 201, in send_messages
    dkim_headers=self.dkim_headers)
  File "…/Vcs/django-ses/django_ses/__init__.py", line 42, in dkim_sign
    include_headers=dkim_headers)
  File "…/Vcs/dkimpy/dkim/__init__.py", line 1159, in sign
    return d.sign(selector, domain, privkey, identity=identity, canonicalize=canonicalize, include_headers=include_headers, length=length)
  File "…/Vcs/dkimpy/dkim/__init__.py", line 720, in sign
    raise ParameterError("The From header field MUST be signed")
dkim.ParameterError: The From header field MUST be signed
```

This is caused, since on py3 `b"foo" in ["foo"]` is False, while it is True on py2.

I think the check should use normal strings, but I see a lot of byte strings being used in the code altogether.

Is `include_headers` meant to be a list of bytes?
(there does not seem to be any documentation, is there?)

[Scott Kitterman (kitterman)]

We use b"foo" throughout dkimpy because getting comparisons to work in both python2 and python3 both is harder otherwise.

It's true that `b"foo" in ["foo"]` is False in python3, but`b"foo" in [b"foo"]` is true.

So yes, bytes.

As far as documentation goes, there isn't much. I'm open to suggestions on how best to document this.

[Scott Kitterman (kitterman)]

I'm looked into this and confirmed it's not reasonable to change this. I've retitled the bug to be about documenting it. I'll do something with that.

[Scott Kitterman (kitterman)]

Actually, thought about it some more. We can handle this case.

[Scott Kitterman (kitterman)]

Fixed in 0.8.1.

[Daniel Hahler (blueyed)]

Thank you for fixing this!

Without checking I think there are other settings affected as well.

I think they should get fixed in the same way then.

(I am not using the lib myself, but only came across this with django-ses's tests - which currently still use pydkim).

[Scott Kitterman (kitterman)]

On Sunday, June 17, 2018 04:23:28 PM you wrote:
> Without checking I think there are other settings affected as well.

There are. Please open a new bug for that and I'll look into it.

Scott K