ASCII tostring corrupts Russian data

Parsing Russian XML content, then serializing as an ASCII string produces corrupted data. A paragraph is truncated, and ends with `\x10/p>`. Using encoding="utf8" or encoding="unicode" avoids the problem.

To reproduce:

---- 8< --------------
import requests
from lxml import etree

URL = "https://raw.githubusercontent.com/nedbat/nedbatcom/a53aa4d2a4aff80cad8775700b2c4866fd2cc795/pages/text/deleting-code_ru.px"
FILENAME = "data.px"

# Get the bytes in a file and parse them to DOM.
with open(FILENAME, "wb") as f:
    f.write(requests.get(URL).content)
element = etree.parse(FILENAME).getroot()

# An ascii string has corrupted data.
text = etree.tostring(element).decode('utf8')
print("\x10/p>" in text)

# A UTF8 string is not corrupted.
text = etree.tostring(element, encoding="utf8").decode("utf8")
print("\x10" in text)

# A Unicode string is not corrupted.
text = etree.tostring(element, encoding="unicode")
print("\x10" in text)
-----------------------------

Versions:
Python : sys.version_info(major=3, minor=9, micro=7, releaselevel='final', serial=0)
lxml.etree : (4, 6, 3, 0)
libxml used : (2, 9, 10)
libxml compiled : (2, 9, 10)
libxslt used : (1, 1, 34)
libxslt compiled : (1, 1, 34)