fix(benchmarks): unblock codspeed runs on pytest 9

[collerek]

Summary

• Drop the no-op @pytest.mark.benchmark(...) decorator on the aio_benchmark fixture in benchmarks/conftest.py. Marks on fixtures never had any effect and pytest 9 promoted the deprecation to a hard PytestRemovedIn9Warning raised at conftest import time, so every benchmark run since the pytest 8.4.2 → 9.0.3 bump (build(deps-dev): bump pytest from 8.4.2 to 9.0.3 #1609) hung and got auto-cancelled at the 24h job timeout.
• Prefix the rewritten name with "u_" in test_benchmark_update.py and test_benchmark_bulk_update.py. The assertion author.name != starting_first_name previously compared two random 5-char strings drawn from string.ascii_letters — over thousands of codspeed walltime iterations the ~1/311M per-call collision eventually surfaced (CI showed 'DpwEX' != 'DpwEX'). The prefix makes the new value structurally unable to match.

Without this, master has had no successful benchmark baseline since 53d2462 (2026-03-31), which is why the recent feature/flatten-fields PR (#1641) shows a phantom -35% regression and a "Different runtime environments detected" warning — codspeed is comparing against a 5-week-old run on different macro-runner hardware.

Test plan

• pytest benchmarks/test_benchmark_update.py benchmarks/test_benchmark_bulk_update.py --codspeed passes locally on pytest 9.0.3 + pytest-codspeed 4.5.0
• codspeed-benchmarks CI succeeds on this PR, producing a fresh master baseline once merged

[codspeed-hq]

Merging this PR will degrade performance by 34.64%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 2 improved benchmarks
❌ 49 (👁 49) regressed benchmarks
✅ 33 untouched benchmarks

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
👁	WallTime	test_count[250]	3 ms	3.6 ms	-16.6%
👁	WallTime	test_avg[500]	2.5 ms	3.4 ms	-25.65%
👁	WallTime	test_max[250]	2.4 ms	3.4 ms	-27.16%
👁	WallTime	test_making_and_inserting_models_in_bulk[10]	9.5 ms	10.6 ms	-10.62%
👁	WallTime	test_get_or_none[250]	2.6 ms	3.5 ms	-25.38%
👁	WallTime	test_get_or_create_when_create[10]	102.4 ms	119.7 ms	-14.44%
👁	WallTime	test_get_or_create_when_create[40]	413.6 ms	484.2 ms	-14.57%
👁	WallTime	test_saving_models_individually[20]	149.3 ms	167.6 ms	-10.91%
👁	WallTime	test_values[250]	5.5 ms	6.4 ms	-14.2%
👁	WallTime	test_min[500]	2.5 ms	3.4 ms	-25.16%
👁	WallTime	test_first[500]	2.5 ms	3.3 ms	-25.67%
👁	WallTime	test_get_or_none[1000]	2.7 ms	3.5 ms	-24.14%
👁	WallTime	test_deleting_all[250]	1.4 ms	2.2 ms	-34.64%
👁	WallTime	test_avg[250]	2.4 ms	3.5 ms	-30.88%
👁	WallTime	test_get_or_create_when_create[20]	209.4 ms	240.7 ms	-13.01%
👁	WallTime	test_get_one[500]	2.7 ms	3.5 ms	-24.54%
👁	WallTime	test_get_or_none[500]	2.7 ms	3.5 ms	-24.56%
👁	WallTime	test_values_list[250]	5.6 ms	6.5 ms	-13.53%
👁	WallTime	test_get_or_create_when_get[1000]	2.7 ms	3.5 ms	-24.48%
👁	WallTime	test_get_one[250]	2.7 ms	3.5 ms	-24.18%
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

---

_{Comparing fix/benchmarks-pytest9 (695f4ea) with master (53d2462)}