gh-156031: Restore urllib.request._opener in test_httpservers

[ayaangazali]

test_httpservers leaves urllib.request._opener installed for the rest of the test run, so regrtest flags it as altering the execution environment. Details and the trace are in gh-156031.

Short version: CommandLineRunTimeTestCase.fetch_file() calls urlopen() with no SSL context for the plain HTTP cases, and urlopen() only builds a throwaway opener when a context is passed. Otherwise it assigns the module global:

elif _opener is None:
    _opener = opener = build_opener()

test_urllib already guards the same global with self.addCleanup(urllib.request.urlcleanup) in four places, and urlcleanup() resets _opener, so this uses the same pattern rather than inventing a new one.

Verifying

Repro takes about two seconds. test_urllib runs first only to get urllib.request into sys.modules, which is what makes regrtest watch the resource at all:

./python.exe -m test test_urllib test_httpservers

Before:

Warning -- urllib.requests._opener was modified by test_httpservers
test_httpservers failed (env changed)

After: SUCCESS, same 209 tests run.

Also ran the whole suite sequentially, which is the only mode that detects this, before and after:

	before	after
tests OK	459	460
env changed	1 (test_httpservers)	0
failed	1 (test_struct)	1 (test_struct)

test_httpservers goes from failed (env changed) to passed. The test_struct failure is unrelated and present before my change too, so I left it alone. pre-commit passes on the touched file.

Worth noting why this has not turned up in CI: save_env.py only watches the resource when urllib.request is already imported, because try_get_module() raises SkipTestEnvironment otherwise. Under -j each test file gets a fresh subprocess, so the module is not in sys.modules when regrtest snapshots the environment and the resource is never tracked. Running test_httpservers on its own is clean for the same reason. It only shows up sequentially, after something earlier has imported urllib.request.

No test added. The fix is test-only hygiene and the suite run above is the check, so a new test would just restate it.

apologies if any of this is off. I found it by running the suite sequentially, traced the cause myself, and talked the fix choice over with Claude Code before settling on matching the existing test_urllib pattern. happy to be corrected on any of it. freshman in college, doing what I can to help out :)

[bedevere-app]

Most changes to Python require a NEWS entry. Add one using the blurb_it web app or the blurb command-line tool.

If this change has little impact on Python users, wait for a maintainer to apply the skip news label instead.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[aisk]

Test only changes don't requires a news entry file.

[ayaangazali]

Good catch, thanks. Removed the NEWS entry, the diff is back to just the one-line cleanup in the test.

For context on why it was there: bedevere/news went red before the skip news label existed on the PR, and I added a blurb to clear it rather than waiting. Should have left it alone and let a maintainer decide, since that label is not something I can apply. Noted for next time.

[aisk]

Introduced in 3.15, so only need to backport to 3.15 branch.