Add codespeed for continuous (and sudoku)

While pubgrub's performance is critical for both uv and cargo, there currently no benchmarking happening in the repository. [Codspeed](https://codspeed.io) runs our benchmarks with instruction counting, reporting more numbers with less variance than wall time.

We get feedback on every PR, can see trends over time and there are flamegraphs and flamegraph diffs in the web view. We've made good experiences with it in both ruff and uv.

With codspeed installed, we can start adding real-world benchmarks for uv and cargo to the pubgrub repo, and then optimize those.

.github/workflows/benchmarks.yml

@@ -0,0 +1,29 @@
+name: Benchmarks (CodSpeed)
+
+on:
+  push:
+    branches:
+      - main
+  pull_request:
+  workflow_dispatch:
+
+jobs:
+  benchmarks:
+    name: Run benchmarks
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Setup rust toolchain, cache and cargo-codspeed binary
+        uses: moonrepo/setup-rust@v1
+        with:
+          channel: stable
+          cache-target: release
+          bins: cargo-codspeed
+
+      - name: Build the benchmark target(s)
+        run: cargo codspeed build
+
+      - name: Run the benchmarks
+        uses: CodSpeedHQ/action@v3
+        with:
+          run: cargo codspeed run

Cargo.toml

@@ -32,7 +32,7 @@ thiserror = "1.0"
 version-ranges = { version = "0.1.0", path = "version-ranges" }
 
 [dev-dependencies]
-criterion = "0.5"
+criterion = { version = "2.7.2", package = "codspeed-criterion-compat" }
 env_logger = "0.11.5"
 proptest = "1.5.0"
 ron = "=0.9.0-alpha.0"
@@ -46,3 +46,7 @@ serde = ["dep:serde", "version-ranges/serde"]
 name = "large_case"
 harness = false
 required-features = ["serde"]
+
+[[bench]]
+name = "sudoku"
+harness = false