Incorrect decoding of preamble in email parser

Bug report

Bug description:

Method: email.message_from_binary_file()

It seems we've found an issue with the email.parser module when parsing raw binary MIME message where the preamble contains UTF-8 encoded data

When using the .as_string() method on the returned message, the unicode data will contain invalid UTF-8 characters ("surrogates not allowed")
The problem does not occur when using the email.message_from_string() method

I've made a somewhat minimal example (from the MIME RFC) exposing the issue.
Here "préamble" is decoded as "pr\udcc3\udca9amble"

import io
import email
import email.policy

CONTENTS = """
From: Nathaniel Borenstein <nsb@bellcore.com>
To: Ned Freed <ned@innosoft.com>
Subject: Sample message
MIME-Version: 1.0
Content-type: multipart/mixed; boundary="i-am-boundary"

This is the préamble.  It is to be ignored, though it
is a handy place for mail composers to include an
explanatory note to non-MIME compliant readers.

--i-am-boundary
Content-type: text/plain; charset=us-ascii

This is explicitly typed plain ASCII text.
It DOES end with a linebreak.

--i-am-boundary
Content-type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit

This should be correctly encapsulated: Un petit café ?

--i-am-boundary--
This is the epilogue.  It is also to be ignored.

""".lstrip()
CONTENTS_BYTES = io.BytesIO(CONTENTS.encode())

# Does not have an impact on the result
POLICY = email.policy.default.clone(utf8=True)


def show_message(msg):
    # Parts are correctly decoded in all cases
    for i, part in enumerate(msg.iter_parts(), 1):
        print(f'MIME PART {i}:')
        as_string = part.as_string()
        as_bytes = as_string.encode()
        print(as_string)

    as_string = msg.as_string()

    # When source was bytes, the unicode result of as_string is incorrect
    as_bytes = as_string.encode()
    print(as_string)


msg_from_binary = email.message_from_binary_file(CONTENTS_BYTES, policy=POLICY)
show_message(msg_from_binary)
# UnicodeEncodeError: 'utf-8' codec can't encode characters in position 192-193: surrogates not allowed

# Using the unicode representation is OK
#msg_from_string = email.message_from_string(CONTENTS, EmailMessage)
#show_message(msg_from_string)

I've tried adding a charset and content-transfer-encoding: 8bit in the headers, with the same result (I do not know if this is actually valid)

Looking at the current code it seems that BytesParser always uses the ASCII encoding with errors='surrogateescape'

CPython versions tested on:

3.11, CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-134384

Start with the minimal binary MIME example in the issue and inspect Lib/email/parser.py around lines 95-103, comparing binary-file parsing with string parsing. Check linked PR gh-134384 for the current work. Done means UTF-8 data in the preamble no longer becomes surrogate characters when the parsed message is rendered.