OnPair: drop Like pushdown for now, keep Compare token-aware path

The byte-streaming `prefix%` and per-row decode + memmem `%contains%`
implementations were not consistently faster than canonicalize + scalar
LIKE: the bulk 4×-unrolled decoder is hard to beat with per-row work.
Drop Like from PARENT_KERNELS so the system falls through to
canonicalize + scalar LIKE.

Compare stays pushed: LPM-tokenise the literal once, then `&[u16]`
equality on every row's `codes[lo..hi]` — no decode at all, ~7 ns/row.

Tests still pass via the canonicalize fallback. A token-DFA
implementation (FSST-style, EQSearch / PrefixAutomaton on tokens) is
tracked for the next iteration.

Signed-off-by: claude <claude@anthropic.com>

Author: claude

encodings/onpair/src/kernel.rs

@@ -5,13 +5,17 @@ use vortex_array::arrays::filter::FilterExecuteAdaptor;
 use vortex_array::kernel::ParentKernelSet;
 use vortex_array::scalar_fn::fns::binary::CompareExecuteAdaptor;
 use vortex_array::scalar_fn::fns::cast::CastExecuteAdaptor;
-use vortex_array::scalar_fn::fns::like::LikeExecuteAdaptor;
 
 use crate::OnPair;
 
+// Compare is pushed: LPM-tokenise the literal once, compare the row's
+// `codes[lo..hi]` against the token sequence as `&[u16]` — no decode.
+// Like is currently *not* registered: the per-row byte-streaming /
+// `memmem`-on-decoded-row implementations are slower than letting the
+// canonicalize + scalar `LIKE` path run. A token-DFA pushdown (FSST-
+// style) is the right replacement and tracked as future work.
 pub(super) const PARENT_KERNELS: ParentKernelSet<OnPair> = ParentKernelSet::new(&[
     ParentKernelSet::lift(&CastExecuteAdaptor(OnPair)),
     ParentKernelSet::lift(&CompareExecuteAdaptor(OnPair)),
     ParentKernelSet::lift(&FilterExecuteAdaptor(OnPair)),
-    ParentKernelSet::lift(&LikeExecuteAdaptor(OnPair)),
 ]);