perf: LIKE/GLOB prefix range, BETWEEN single-range, and IN-list index optimizations

[iheitlager]

Summary

Carried forward from #527 (epic #421, closed without full completion).
Verified via codebase search: LIKE/GLOB prefix-to-index-range rewriting has
zero matches anywhere in src/. BETWEEN handling exists in expression
codegen (src/codegen/select/aggregate.rs, src/codegen/select/join_order.rs)
but it's unconfirmed whether it collapses to a single index range seek
versus two separate constraints. IN-list-to-multiple-index-seeks is also
unverified.

Techniques

Technique	Description	Priority
LIKE/GLOB prefix	LIKE 'foo%' → index range >= 'foo' AND < 'fop'	High
BETWEEN → range	x BETWEEN 5 AND 10 → x >= 5 AND x <= 10 with single index seek	Medium
IN list optimization	x IN (1,2,3) → multiple index seeks	Medium

Acceptance criteria

• EXPLAIN QUERY PLAN shows index usage for LIKE 'prefix%'
• BETWEEN uses a single index range seek, not two separate constraints
• x IN (...) against an indexed column uses one seek per value, not a
  full scan
• Benchmark shows improvement for pattern-matching and range queries

Complexity

Estimate: medium
Reasoning: LIKE prefix extraction needs a literal-prefix analysis of the
pattern (bail out on leading wildcard/escape edge cases) plus wiring into
the planner's range-constraint machinery — new code path. BETWEEN/IN need
verification first (may already partially work) and then closing gaps in
src/planner.rs / index-seek codegen if they don't collapse as expected.

Refs: #527