fix(get_addresses): accept email.header.Header inputs (no more 'Header' object has no attribute 'strip')

[rakurtz]

I ran into a .eml file validation error when an email contained certain strings / encodings in the header. I could be tracked down to a call of .strip() on a non-str object.

Summary

mailparser.utils.get_addresses(raw_header) calls raw_header.strip() and
passes raw_header to email.utils.getaddresses([raw_header], strict=True).
Both assume a plain str, but core.py feeds get_addresses whatever
email.message.Message.get(name) returns:

# src/mailparser/core.py
elif name_header in ADDRESSES_HEADERS:
    raw_header = self.message.get(name_header, "") if self.message else ""
    ...
    parsed_addresses = get_addresses(raw_header)

For any address header containing RFC 2047 encoded-words — typical for non-ASCII display names like =?utf-8?q?=C3=81rp=C3=A1d?= user@example.com — Message.get() may return an email.header.Header instance. That instance has no .strip(), so MailParser.from_bytes(...) (or parse_from_bytes) raises:

AttributeError: 'Header' object has no attribute 'strip'
…on perfectly valid mail and the entire parse fails.

This is the same shape of bug that #149 fixed in ported_string; that PR covered attachment header values but the address-parsing path was missed.

Fix

Normalise raw_header once at the top of get_addresses:

if raw_header is None:
        return []
    if isinstance(raw_header, email.header.Header):
        raw_header = decode_header_part(raw_header.encode())
    elif not isinstance(raw_header, str):
        raw_header = str(raw_header)

(edited to reflect the last commit that ensures correct encoding/decoding)

email.header.Header.__str__ already produces the encoded-word string form that the existing _ADDR_FALLBACK_RE and email.utils.getaddresses know how to handle, and core.py decodes the display name afterwards via decode_header_part. The strict-string happy path is unchanged.

Tests

Adds three focused unit tests in tests/test_utils.py::TestUtilsEdgeCases (matching the style of the existing test_ported_string_handles_header_object test from #149):

test_get_addresses_handles_header_object — direct regression. Constructs an email.header.Header from an encoded-word value and asserts get_addresses returns the parsed [(display_name, "user@example.com")] without raising.
test_get_addresses_handles_none — defensive: get_addresses(None) returns [] instead of crashing on attribute access.
test_get_addresses_plain_string_unchanged — guards the existing happy path so the new isinstance(raw_header, str) branch can't silently regress the common case.

Full local test run on Python 3.13:

============================= 190 passed in 1.00s ==============================
TOTAL: 99% coverage
Reproducer (without the fix)
from email.header import Header
from mailparser.utils import get_addresses
get_addresses(Header(
    "=?utf-8?q?=C3=81rp=C3=A1d?= <user@example.com>",
    charset="utf-8",
))
# AttributeError: 'Header' object has no attribute 'strip'

Notes

No behaviour change for str inputs.
Public docstring of get_addresses updated to document the accepted input types (str | email.header.Header | None).
No changes to core.py — the fix is local to the helper so any other callers (current or future) get the same robustness automatically.

[rakurtz]

after the first commit the PR is based on i realized that the string encoding was broken. the second and third commit fix this by decoding the Header values before address parsing.