UTF-8 passphrases fail with latin-1 encoding exception during encryption

[Neverous]

Describe the bug

When using python-gnupg with Deja Dup / Duplicity (which switched to this library recently in version 3.2.0), passphrases containing non-Latin-1 characters or Unicode symbols fail with: Thread exception: 'latin-1' codec can't encode character '***' in position *: ordinal not in range(256)

This occurs because self.encoding hardcodes latin-1: https://github.com/vsajip/python-gnupg/blob/master/gnupg.py#L1155

The only way to override it is externally on the instance, but it's unclear if that is the intended solution. I noticed an old comment suggesting an external override is expected, but shouldn't it be configurable via constructor arguments?

To Reproduce

import gnupg

gpg = gnupg.GPG()
# Passphrase containing non-Latin-1 / Unicode characters
status = gpg.encrypt("Test payload", recipients=None, symmetric=True, passphrase="błąd")
print(status.status)

---------------------------------------------------------------------------
UnicodeEncodeError                        Traceback (most recent call last)

...
UnicodeEncodeError: 'latin-1' codec can't encode characters in position 1-2: ordinal not in range(256)

Expected behavior

python-gnupg should gracefully handle utf-8 encoded non-Latin-1 inputs and passphrases without failing on latin-1 encoding limitations.

Screenshots

N/A

Environment

• OS, including version: Arch Linux
• Version of this library: 0.5.6
• Version of GnuPG: gpg (GnuPG) 2.4.9

Additional information

• Manually setting gpg.encoding = 'utf-8' immediately after GPG instantiation seems to resolve the exception
• Duplicity report for context: https://gitlab.com/duplicity/duplicity/-/work_items/941