Add benchmarking test suite and greatly improve performance in a few cases (#948)

* Add benchmarking test suite

* Improve amortized time of model relation loads with a large number of rows

* Improve performance of loading models with many related models

* Improve performance of loading models with many related models to O(N)ish

* Fix bug where N model creation with shared related model would build in N^2 time

* Lower blocking time for queryset results

* Add docstrings and streamline hash code

Co-authored-by: haydeec1 <Eric.Haydel@jhuapl.edu>

benchmarks/test_benchmark_delete.py

@@ -0,0 +1,36 @@
+from typing import List
+
+import pytest
+
+from benchmarks.conftest import Author
+
+pytestmark = pytest.mark.asyncio
+
+
+@pytest.mark.parametrize("num_models", [250, 500, 1000])
+async def test_deleting_all(
+    aio_benchmark, num_models: int, authors_in_db: List[Author]
+):
+    @aio_benchmark
+    async def delete_all():
+        await Author.objects.delete(each=True)
+
+    delete_all()
+
+    num = await Author.objects.count()
+    assert num == 0
+
+
+@pytest.mark.parametrize("num_models", [10, 20, 40])
+async def test_deleting_individually(
+    aio_benchmark, num_models: int, authors_in_db: List[Author]
+):
+    @aio_benchmark
+    async def delete_one_by_one(authors: List[Author]):
+        for author in authors:
+            await Author.objects.filter(id=author.id).delete()
+
+    delete_one_by_one(authors_in_db)
+
+    num = await Author.objects.count()
+    assert num == 0