Try running benchmarks in parallel and in Python 3.12

.github/workflows/CI.yml

@@ -36,7 +36,7 @@ jobs:
           python-version: ${{ matrix.py }}
           cache: "pip"
       - run: pip install -e .[test]
-      - run: pytest --benchmark-disable
+      - run: pytest --benchmark-disable -vvv
   mypy:
     runs-on: ubuntu-latest
     steps:
@@ -58,13 +58,13 @@ jobs:
       - uses: Swatinem/rust-cache@v2
       - uses: actions/setup-python@v4
         with:
-          python-version: "3.10"
+          python-version: "3.12"
           cache: "pip"
       - run: pip install -e .[test]
       - uses: CodSpeedHQ/action@v2
         with:
           token: ${{ secrets.CODSPEED_TOKEN }}
-          run: pytest -vvv
+          run: pytest -vvv -n auto
   docs:
     runs-on: ubuntu-latest
     steps:

pyproject.toml

@@ -39,6 +39,7 @@ test = [
     "egglog[array]",
     "pytest-codspeed",
     "pytest-benchmark",
+    "pytest-xdist",
 ]
 
 docs = [

python/tests/test_array_api.py

@@ -114,7 +114,7 @@ def load_source(expr):
     return egraph.eval(fn_program.statements)
 
 
-@pytest.mark.benchmark(min_rounds=1)
+@pytest.mark.benchmark(min_rounds=3)
 class TestLDA:
     def test_trace(self, snapshot_py, benchmark):
         @benchmark