chore: disable SQLite disk sync in benchmarks to reduce walltime noise

[collerek]

Benchmarks that write rows one by one were dominated by SQLite fsync and journal-file I/O, so CodSpeed's walltime runs flagged random 15–30% "regressions" on PRs that never touch those code paths (e.g. #1840, #1824). The benchmark fixture now sets PRAGMA synchronous=OFF and journal_mode=MEMORY on SQLite connections, so these benchmarks measure ormar overhead instead of disk latency.

Locally, over 3 runs of the 21 write-per-row benchmarks, the mean run-to-run spread dropped from 17% to 6% (max 39% → 12%). Expect the first CodSpeed report here to show large improvements on those benchmarks, as the baseline resets once this is merged.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[codspeed]

Merging this PR will improve performance by ×2.5

⚡ 39 improved benchmarks
✅ 84 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	test_updating_models_individually[40]	323.3 ms	106.9 ms	×3
⚡	test_saving_models_individually[20]	157.5 ms	52.1 ms	×3
⚡	test_saving_models_individually[40]	314.9 ms	104.7 ms	×3
⚡	test_updating_models_individually[10]	80.6 ms	26.9 ms	×3
⚡	test_updating_models_individually[20]	161.1 ms	53.8 ms	×3
⚡	test_creating_models_individually[40]	318.4 ms	107 ms	×3
⚡	test_creating_models_individually[20]	158.8 ms	53.5 ms	×3
⚡	test_saving_models_individually[10]	78.2 ms	26.4 ms	×3
⚡	test_update_or_create_when_create[40]	318.3 ms	107.8 ms	×3
⚡	test_creating_models_individually[10]	79.3 ms	26.9 ms	×2.9
⚡	test_update_or_create_when_create[20]	158.9 ms	54 ms	×2.9
⚡	test_saving_related_new_children[10]	92.2 ms	31.7 ms	×2.9
⚡	test_update_or_create_when_create[10]	77.9 ms	26.9 ms	×2.9
⚡	test_saving_related_new_children[20]	177.4 ms	61.3 ms	×2.9
⚡	test_saving_models_individually_with_related_models[40]	339.5 ms	118.8 ms	×2.9
⚡	test_saving_models_individually_with_related_models[20]	169.5 ms	59.4 ms	×2.9
⚡	test_saving_related_new_children[40]	341.9 ms	120.3 ms	×2.8
⚡	test_adding_new_children_to_reverse_relation[40]	340 ms	119.7 ms	×2.8
⚡	test_saving_models_individually_with_related_models[10]	84.7 ms	29.8 ms	×2.8
⚡	test_adding_new_children_to_reverse_relation[20]	168.4 ms	60.1 ms	×2.8
...	...	...	...	...

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

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing chore/benchmarks-sqlite-no-sync (30bbbb8) with master (8993d4e)}