Improve FSST LIKE contains handling

[gatesn]

Rational for this change

FSST's compressed DFA path is not always the fastest path for short %needle% LIKE patterns. For short substring needles, decoding the FSST array once and using the existing canonical LIKE implementation is faster than walking the compressed byte stream through the DFA.

This also makes Shared transparent to parent reductions so FSST parent kernels are still reached when arrays are wrapped by the shared cache layer.

What changes are included in this PR?

Short contains-style FSST LIKE patterns now fall back to canonicalized LIKE evaluation. The FSST LikeKind parser is made crate-visible so the LIKE kernel can choose that path before constructing the DFA matcher.

The PR also adds a regression test covering FSST parent kernels through SharedArray.

What APIs are changed? Are there any user-facing changes?

No public API changes. The behavior change is internal query execution: matching results are unchanged, but some FSST LIKE predicates should execute faster.

[codspeed-hq]

Merging this PR will degrade performance by 36.13%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚡ 2 improved benchmarks
❌ 9 regressed benchmarks
✅ 1578 untouched benchmarks
⏩ 4 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	fsst_contains[path]	5 ms	10.9 ms	-54.35%
❌	Simulation	fsst_contains[email]	4.4 ms	9.5 ms	-53.28%
❌	Simulation	like_substr_high_match	8.1 ms	16.6 ms	-51%
❌	Simulation	fsst_contains[json]	13.9 ms	28.3 ms	-50.99%
❌	Simulation	fsst_contains[cb]	15.5 ms	30.4 ms	-49.12%
❌	Simulation	chunked_bool_canonical_into[(1000, 10)]	16.3 µs	26.9 µs	-39.46%
❌	Simulation	fsst_contains[log]	25.4 ms	40.5 ms	-37.35%
❌	Simulation	fsst_contains[urls]	10.4 ms	15 ms	-30.87%
❌	Simulation	slice_empty_vortex	310 ns	368.3 ns	-15.84%
⚡	Simulation	bitwise_not_vortex_buffer_mut[128]	244.4 ns	215.3 ns	+13.55%
⚡	Simulation	bitwise_not_vortex_buffer_mut[1024]	304.7 ns	275.6 ns	+10.58%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing ngates/fsst-like-pushdown (d15418f) with develop (2a19323)}

Footnotes

1. 4 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩