Execute tensor product functions with RowFn (#9769)

## Summary

- Tracking Issue: #9128
- Depends on: #9768

Moves inner product and cosine similarity to RowFn without changing
their arithmetic contracts.

## Changes

Inner product keeps the existing left-to-right multiply-and-sum order.
Cosine computes that dot product followed by both L2 norms and the
existing zero-denominator guard, with no encoded reduction,
reassociation, or constant-norm shortcut. Tests cover nulls, empty and
zero-width rows, IEEE overflow and underflow, and bitwise agreement
between constant encodings and materialized rows.

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

Author: connortsui20

vortex-tensor/src/scalar_fns/arithmetic.rs

@@ -4,7 +4,8 @@
 //! Row-level arithmetic shared by tensor scalar functions.
 //!
 //! These kernels preserve the scalar functions' left-to-right floating-point arithmetic. Keeping
-//! the arithmetic here ensures that fused and single-result operations use the same result contract.
+//! the arithmetic here ensures that fused and single-result operations use the same result
+//! contract.
 
 use num_traits::Float;
 use vortex_array::dtype::NativePType;
@@ -21,3 +22,14 @@ pub(crate) fn l2_norm_row<T: Float + NativePType>(row: &[T]) -> T {
 
     sum_squared.sqrt()
 }
+
+/// Computes `sum(lhs_i * rhs_i)` for one pair of rows.
+///
+/// Both rows must have the same length. Otherwise, trailing values are ignored and the result is
+/// incorrect. Callers preserve the left-to-right multiplication and accumulation order.
+pub(crate) fn inner_product_row<T: Float + NativePType>(lhs: &[T], rhs: &[T]) -> T {
+    lhs.iter()
+        .zip(rhs)
+        .map(|(&lhs_value, &rhs_value)| lhs_value * rhs_value)
+        .fold(T::zero(), |sum, product| sum + product)
+}