perf: share pk-only related models across rows of a query

[collerek]

Foreign keys that are not selected built a separate pk-only related model for every row (1000 Author placeholders for 1000 books by 50 authors), about half of the time of a plain all(). Rows pointing to the same pk from the same relation now reuse one instance through the per-query row registry from #1851; models below a many-to-many relation stay scoped to their parent.

• Book.objects.all() over 1000 rows is 1.8× faster (179ms → 98ms), reverse select_related 218ms → 149ms; serialized output is byte-identical to master.
• Breaking change (extends the unreleased perf: build each related model once per select_related result #1851 note): chapters[0].book is chapters[1].book for chapters of the same book, the placeholder's reverse relation lists all loaded rows, and await chapter.book.load() fills it in for every row sharing it.
• Also adds the missing perf: run the SQLite engine in AUTOCOMMIT by default #1852 link in the release notes.

[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 regress 2 benchmarks

⚠️ 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_min[500]
• test_avg[250]
• test_get_all_with_related_models[40]
• test_get_all_with_related_models[20]
• test_duplicate_relation_paths[200]
• test_duplicate_relation_paths[100]
• test_duplicate_relation_paths[50]
• test_get_all_with_related_models[10]
• test_avg[500]
• test_creating_wide_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

⚡ 6 improved benchmarks
❌ 2 regressed benchmarks
✅ 118 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌ ⚠️	test_min[500]	5.3 ms	7.2 ms	-26.77%
❌ ⚠️	test_avg[250]	6.4 ms	7.3 ms	-12.66%
⚡ ⚠️	test_get_all_with_related_models[40]	30.6 ms	22 ms	+38.84%
⚡ ⚠️	test_get_all_with_related_models[20]	18.4 ms	14.4 ms	+28.06%
⚡ ⚠️	test_duplicate_relation_paths[200]	259.9 ms	217.4 ms	+19.54%
⚡ ⚠️	test_duplicate_relation_paths[100]	135.3 ms	114.1 ms	+18.61%
⚡ ⚠️	test_duplicate_relation_paths[50]	78.2 ms	67.7 ms	+15.58%
⚡ ⚠️	test_get_all_with_related_models[10]	16.9 ms	14.8 ms	+13.68%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing perf/share-pk-only-related (274c351) with master (67d8106)}