perf: run the SQLite engine in AUTOCOMMIT by default

[collerek]

On SQLite every standalone query switched its pooled connection into AUTOCOMMIT and back through an engine view, costing two hops to the aiosqlite driver thread per query. The SQLite engine now runs in AUTOCOMMIT like MySQL (#1811), and standalone queries use the engine directly when it already is.

• Standalone get / create / save / delete / count are 25–45% faster on SQLite; transactional paths are unchanged.
• SQLite transactions keep their explicit BEGIN, which opens a real transaction on an AUTOCOMMIT connection, so savepoints and rollbacks work without an isolation switch. MySQL still switches to the server default isolation, so the Upgrade from 0.21 to 0.23 degraded our app performance significantly #1629 savepoint error can't come back.
• Postgres is unchanged (no measurable gain), and a user-supplied isolation_level is respected.
• test_standalone_query_uses_autocommit now checks autocommit via _is_autocommit_isolation() and that rows written in a transaction stay invisible to other connections until commit.

[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 26.25%

⚠️ 112 benchmarks spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

• test_deleting_individually[40]
• test_deleting_individually[20]
• test_deleting_individually[10]
• test_adding_saved_children_to_reverse_relation[40]
• test_updating_models_individually[40]
• test_adding_saved_children_to_reverse_relation[20]
• test_updating_models_individually[20]
• test_update_or_create_when_create[40]
• test_saving_models_individually[40]
• test_creating_models_individually[40]
• …and 102 more

⚠️ 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

⚡ 70 improved benchmarks
✅ 56 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡ ⚠️	test_deleting_individually[40]	98.6 ms	60.5 ms	+62.98%
⚡ ⚠️	test_deleting_individually[20]	50.3 ms	31.2 ms	+61.02%
⚡ ⚠️	test_deleting_individually[10]	26.1 ms	16.6 ms	+57.37%
⚡ ⚠️	test_adding_saved_children_to_reverse_relation[40]	237.4 ms	160.1 ms	+48.28%
⚡ ⚠️	test_updating_models_individually[40]	115.7 ms	78 ms	+48.2%
⚡ ⚠️	test_adding_saved_children_to_reverse_relation[20]	120.4 ms	81.9 ms	+47.06%
⚡ ⚠️	test_updating_models_individually[20]	58.9 ms	40.2 ms	+46.55%
⚡ ⚠️	test_update_or_create_when_create[40]	117.2 ms	80.6 ms	+45.45%
⚡ ⚠️	test_saving_models_individually[40]	115.4 ms	79.4 ms	+45.32%
⚡ ⚠️	test_creating_models_individually[40]	116.8 ms	80.4 ms	+45.31%
⚡ ⚠️	test_updating_models_individually[10]	30.7 ms	21.2 ms	+44.89%
⚡ ⚠️	test_get_or_create_when_create[40]	253.8 ms	176.3 ms	+43.97%
⚡ ⚠️	test_saving_models_individually[20]	60.9 ms	42.6 ms	+42.9%
⚡ ⚠️	test_adding_saved_children_to_reverse_relation[10]	64.3 ms	45 ms	+42.89%
⚡ ⚠️	test_creating_models_individually[20]	61.4 ms	43 ms	+42.71%
⚡ ⚠️	test_update_or_create_when_create[20]	61.6 ms	43.2 ms	+42.61%
⚡ ⚠️	test_get_or_create_when_create[20]	130.5 ms	91.9 ms	+42.02%
⚡ ⚠️	test_get_or_create_when_create[10]	70.2 ms	50.8 ms	+37.97%
⚡ ⚠️	test_update_or_create_when_create[10]	33.8 ms	24.6 ms	+37.47%
⚡ ⚠️	test_saving_models_individually[10]	34 ms	24.8 ms	+36.81%
...	...	...	...	...

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 perf/engine-level-autocommit (2247028) with master (e22d2d7)}