Add scalar function performance baselines (#9136)

## Summary

Tracking Issue: #9128

Adds stable scalar function performance baselines for `byte_length`,
`l2_norm`, `normalized`, `inner_product`, `cosine_similarity`,
`GeoContains`, and `GeoIntersects`.

The benchmark sizes are all comfortably above the roughly
146-instruction Divan harness floor from #9011. Each benchmark binary
also uses vendored `mimalloc`, since scalar function execution allocates
its output inside the timed trace and glibc differences between runner
images caused the flakes fixed in #8861.

## Why

Codspeed can only compare a later implementation change when the same
benchmark name already exists on `develop`, so these baselines need to
land before the `RowFn` things I want to make. And these are pretty
cheap, might as well add them.

---------

Signed-off-by: Connor Tsui <connor.tsui20@gmail.com>

Author: connortsui20

.github/workflows/codspeed.yml

@@ -57,6 +57,7 @@ jobs:
           - { shard: 6, name: "Encodings 3", packages: "vortex-pco vortex-runend vortex-sequence" }
           - { shard: 7, name: "Encodings 4", packages: "vortex-sparse vortex-zigzag vortex-zstd" }
           - { shard: 8, name: "Storage formats & row encoding", packages: "vortex-flatbuffers vortex-proto vortex-btrblocks vortex-row" }
+          - { shard: 9, name: "Tensor & geo", packages: "vortex-tensor vortex-geo" }
     name: "Benchmark with Codspeed (Shard #${{ matrix.shard }})"
     timeout-minutes: 30
     runs-on: >-

Cargo.lock

@@ -36,7 +36,7 @@ fn main() {
 
 static SESSION: LazyLock<VortexSession> = LazyLock::new(array_session);
 
-const LEN: usize = 65_536;
+const LEN: usize = 32_768;
 
 #[divan::bench]
 fn add_i64_nonnull(bencher: Bencher) {
@@ -54,6 +54,30 @@ fn add_i64_nullable(bencher: Bencher) {
     bench_primitive(bencher, lhs, rhs, Operator::Add);
 }
 
+#[divan::bench]
+fn add_i64_constant(bencher: Bencher) {
+    let lhs = primitive_nonnull(0).into_array();
+    let rhs = ConstantArray::new(1_000_000i64, LEN).into_array();
+
+    bench_primitive(bencher, lhs, rhs, Operator::Add);
+}
+
+#[divan::bench]
+fn add_i32_nonnull(bencher: Bencher) {
+    let lhs = primitive_i32_small_nonnull(1).into_array();
+    let rhs = primitive_i32_small_nonnull(17).into_array();
+
+    bench_primitive(bencher, lhs, rhs, Operator::Add);
+}
+
+#[divan::bench]
+fn add_u32_nonnull(bencher: Bencher) {
+    let lhs = primitive_u32_small_nonnull(1).into_array();
+    let rhs = primitive_u32_small_nonnull(17).into_array();
+
+    bench_primitive(bencher, lhs, rhs, Operator::Add);
+}
+
 #[divan::bench]
 fn mul_i64_nonnull(bencher: Bencher) {
     let lhs = primitive_small_nonnull(1).into_array();

vortex-array/benches/binary_ops.rs

@@ -36,6 +36,7 @@ _test-harness = []
 
 [dev-dependencies]
 divan = { workspace = true }
+mimalloc = { workspace = true }
 rstest = { workspace = true }
 vortex-array = { workspace = true, features = ["_test-harness"] }
 vortex-geo = { path = ".", features = ["_test-harness"] }
@@ -49,5 +50,13 @@ harness = false
 name = "predicate_bbox"
 harness = false
 
+[[bench]]
+name = "binary_predicates"
+harness = false
+
+[[bench]]
+name = "distance"
+harness = false
+
 [lints]
 workspace = true