Halve listview zip bench LEN to stay under CodSpeed simulation budget

CodSpeed's instruction-count simulation runs ~10x local walltime, putting the
8192-list bench at ~550us there. Drop to 4096 lists so each case stays well
under a few hundred microseconds in CI while still exercising the select.

Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

Author: claude

vortex-array/benches/listview_zip.rs

@@ -21,9 +21,10 @@ fn main() {
 }
 
 // Smaller than the value-path benches: listview zip cost is dominated by element concatenation and
-// per-list canonicalization, so a few thousand lists already exercises the select while keeping the
-// benchmark well under a few hundred microseconds.
-const LEN: usize = 8_192;
+// per-list canonicalization. A few thousand lists already exercise the select while keeping each
+// case well under a few hundred microseconds under CodSpeed's instruction-count simulation, which
+// runs ~10x the local walltime.
+const LEN: usize = 4_096;
 
 /// Fragmented (alternating) mask: the worst case for the per-element branch this kernel replaces.
 /// The branchless chunked select is mask-shape-independent, so one shape suffices.

vortex-array/src/arrays/listview/compute/zip.rs

@@ -450,7 +450,7 @@ mod tests {
             sizes.iter().copied().collect::<Buffer<u64>>().into_array(),
             Validity::NonNullable,
         );
-        assert_arrays_eq!(result, expected);
+        assert_arrays_eq!(result, expected, &mut ctx);
         Ok(())
     }