try pgo for benchmarks run

.github/workflows/codspeed.yml

@@ -41,19 +41,39 @@ jobs:
         if: steps.cache-py.outputs.cache-hit != 'true'
 
       - name: install rust stable
+        id: rust-toolchain
         uses: dtolnay/rust-toolchain@stable
+        with:
+          components: llvm-tools
 
       - name: cache rust
         uses: Swatinem/rust-cache@v2
 
-      - name: Install pydantic-core
+      - name: Compile pydantic-core for profiling
+        # build-prod does not include MiMalloc because it bypasses Maturin
+        run: |
+          pip install -e .
+          make build-prod
+          python -c 'import pydantic_core'
+        env:
+          CONST_RANDOM_SEED: 0 # Fix the compile time RNG seed
+          RUSTFLAGS: "-Cprofile-generate=${{ github.workspace }}/profdata"
+
+      - name: Gather pgo data
+        run: pytest
+
+      - name: Prepare merged pgo data
+        run: ~/.rustup/toolchains/${{ steps.rust-toolchain.outputs.name }}/lib/rustlib/x86_64-unknown-linux-gnu/bin/llvm-profdata merge -o ${{ github.workspace }}/merged.profdata ${{ github.workspace }}/profdata
+
+      - name: Compile pydantic-core for benchmarking
         # build-prod does not include MiMalloc because it bypasses Maturin
         run: |
           pip install -e .
           make build-prod
           python -c 'import pydantic_core'
         env:
           CONST_RANDOM_SEED: 0 # Fix the compile time RNG seed
+          RUSTFLAGS: "-Cprofile-use=${{ github.workspace }}/merged.profdata"
 
       # this is required so that pytest uses the installed package
       - run: rm tests/__init__.py