Quadratic time in `xml.etree.ElementTree` when parsing text with a large number of comments

Bug report

Bug description:

import time
import xml.etree.ElementTree as ET

for N in (5000, 10000, 20000, 40000, 80000):
    data = b"<r>" + b"x<!---->" * N + b"</r>"
    s = time.perf_counter()
    ET.fromstring(data)
    dt = time.perf_counter() - s
    print(f"{N}  {dt}s")

I see:

$ python repro.py 
5000  0.025435873976675794s
10000  0.0888163199997507s
20000  0.3610062320076395s
40000  1.4099939750158228s
80000  5.41402202800964s

Found by OSS-Fuzz.

CPython versions tested on:

CPython main branch

Operating systems tested on:

No response

Linked PRs

• gh-155407

Start with the ET.fromstring() reproducer in the issue and compare timings as the number of comments grows. Review linked PR gh-155407 to understand the work already underway; done means parsing this input no longer shows quadratic-time growth.