simplify_regex() produces corrupted output for URL patterns with multiple unnamed capture groups

[sepuri_sai_krishna]

simplify_regex() re-emits the whole pattern prefix once per capture group, so
any re_path() containing two or more unnamed capture groups is rendered
with a duplicated, malformed path.

Both callers are affected: the admindocs "Views" page
(ViewIndexView.get_context_data() puts simplify_regex(regex) straight into
the url context value) and the new listurls management command added in 6.2.

Reproduction

>>> from django.contrib.admindocs.views import simplify_regex  # django.urls.utils on main
>>> simplify_regex(r'^articles/(\w+)/comments/(\d+)/$')

Expected:

/articles/<var>/comments/<var>/

Actual:

/articles/<var>/comments/articles/(\w)/comments/<var>/

The corruption compounds with each additional group. With three:

>>> simplify_regex(r'^a/(\w+)/b/(\d+)/c/(\w+)$')
'/a/<var>/b/a/(\w)/b/<var>/c/a/(\w)/b/(\d)/c/<var>'   # expected '/a/<var>/b/<var>/c/<var>'

Named groups mixed with unnamed ones are affected too:

>>> simplify_regex(r'^shop/(?P<cat>\w+)/(\w+)/(\d+)/$')
'/shop/<cat>/<var>/shop/<cat>/(\w)/<var>/'            # expected '/shop/<cat>/<var>/<var>/'

End-to-end, via listurls on main

With this URLconf:

urlpatterns = [
    re_path(r"^articles/(\w+)/comments/(\d+)/$", view),
    re_path(r"^shop/(?P<cat>\w+)/(\w+)/(\d+)/$", view),
    re_path(r"^single/(\w+)/$", view),
]

manage.py listurls -f stacked prints:

Route: /articles/<var>/comments/articles/(\w)/comments/<var>/
Route: /shop/<cat>/<var>/shop/<cat>/(\w)/<var>/
Route: /single/<var>/

Only the single-group pattern is correct.

Affected versions

Verified by running the snippet above against released wheels:

Version	Output for ^articles/(\w+)/comments/(\d+)/$
4.2.16	/articles/<var>/comments/articles/(\w)/comments/<var>/
5.2.9	/articles/<var>/comments/articles/(\w)/comments/<var>/
6.0.7	/articles/<var>/comments/articles/(\w)/comments/<var>/
main	/articles/<var>/comments/articles/(\w)/comments/<var>/

This is long-standing rather than a regression; the same construct is present in
the code that predates both the move of these helpers to django/urls/utils.py
in 6.2 and the changes made for #30731.

Cause

In replace_unnamed_groups() (django/urls/utils.py on main;
django/contrib/admindocs/utils.py on 6.0 and earlier):

final_pattern, prev_end = "", None
for start, end, _ignored in _find_groups(pattern, _UNNAMED_GROUP_MATCHER):
    if prev_end:
        final_pattern += pattern[prev_end:start]
    final_pattern += pattern[:start] + "<var>"     # <-- re-emits the prefix
    prev_end = end
return final_pattern + pattern[prev_end:]

On the first iteration prev_end is None and pattern[:start] is correct. On
every later iteration the text between the previous group and this one is
appended (correctly), and then pattern[:start] appends the entire prefix from
index 0 a second time.

The sibling function remove_non_capturing_groups() in the same module already
has the correct shape, relying on pattern[None:start] == pattern[:start] for
the first iteration:

final_pattern += pattern[prev_end:start]

So the fix collapses the three lines into one that matches the existing idiom:

final_pattern += pattern[prev_end:start] + "<var>"

Why this was never caught

SimplifyRegexTests.test_simplify_regex (tests/urlpatterns/tests.py) has over
100 cases, but every one of them contains at most one unnamed group, so
the second-iteration path is never executed.

Patch

PR adds six regression cases covering two and three unnamed groups, unnamed
groups mixed with named ones, nested alternations, and escaped parentheses. All
six fail before the fix and pass after. Full run of urlpatterns,
urlpatterns_reverse, admin_docs, admin_scripts, view_tests and
user_commands is green (733 tests).

A release note is included in docs/releases/6.0.8.txt as the fix qualifies for
backport.