test(project): keep the #1634 scratch repositories' clones off auto-maintenance (#1906)

[JarryShaw]

• Searched for similar pull requests

• Followed the coding style (make pylint, make mypy, make isort)

• make test passes, and a test case covers the change

• Added a changelog entry — N/A — centralised in docs(changelog): shared 1.5.0 changelog — long-lived, merges last (#610, #616, #617, #618, #620) #657

• test — adds or fixes tests

A push to a local path starts receive-pack with GIT_CONFIG_COUNT stripped, so the #1644 maintenance settings never reached origin.git. Its auto-maintenance could then repack while checkout() cloned locally, and the clone failed with ENOENT (git 2.55, builtin/clone.c copy_or_link_directory).

ScratchRepository.setUp now writes a GIT_CONFIG_GLOBAL file before any git command runs. It sets maintenance.auto=false, maintenance.autoDetach=false, gc.auto=0, gc.autoDetach=false and receive.autogc=false, and it replaces the per-class GIT_CONFIG_COUNT overrides.

A new test runs under every subclass and uses GIT_TRACE2_EVENT to assert that no maintenance or gc process starts. It fails without the fix. Under a forced repack on every push, runs went from failing to clean on 2.55.0.

Closes #1906

[JarryShaw]

GOOD TO GO at 474ee695b. A Sonnet cross-review checked the git v2.55.0 source. local_repo_env strips GIT_CONFIG_COUNT from the local push's receive-pack, but not GIT_CONFIG_GLOBAL, and receive.autogc defaults on. On 2.47.3 it confirmed that the old setup still started maintenance run --auto --detach. The new trace2 test fails on every subclass without the fix, asserts that the trace reached receive-pack, and passes with the fix; the file has 85 tests and all pass. The race itself was not reproduced locally, since the local git is 2.47.3. Main's 3.13 re-run of 95caa8b36 passed.

[github-actions]

Coverage: 90.07% (unit tier, Python 3.15, 474ee695b, Unit Tests run success)

Package	Statements	Missed	Branches	Partial	Cover
pcapkit (top level)	104	2	20	2	96.77%
pcapkit/const	19595	881	2618	867	92.11%
pcapkit/corekit	2504	70	792	36	96.54%
pcapkit/dumpkit	276	3	100	3	98.40%
pcapkit/foundation	2916	164	1016	34	93.59%
pcapkit/interface	138	7	52	5	93.68%
pcapkit/protocols	17411	224	4580	189	98.10%
pcapkit/toolkit	872	177	298	8	77.18%
pcapkit/utilities	522	2	158	4	99.12%
pcapkit/vendor	4431	2341	1018	156	43.64%

Per-file detail: the coverage-html artifact of this run.

[JarryShaw]

Further evidence for this fix. #1929's Python 3.15 leg (job 117914902574) hit #1906 again, in TestTheCountsAreOfTheReleasedTiers. This time the clone failed on fatal: failed to copy file to '.../work/.git/objects/maintenance.lock': No such file or directory. That lock file belongs to the auto-maintenance the local push started on origin.git, which is the mechanism this PR switches off. I re-ran the failed jobs on #1929, and its own diff touches no test infrastructure.