perf: pair PostHog raw and DuckLake queries

[bill-ph]

Summary

• migrate the frozen PostHog golden corpus to the paired-query contract from perf: add paired query catalog expansion #1008
• generate identical raw-view and production-shaped DuckLake-table SQL for all seven intents
• alternate pair order on measured iterations: raw/table on odd iterations and table/raw on even iterations
• require an even measurement count for paired catalogs so first-position effects stay balanced
• version all seven query and intent IDs as balanced_v2 so dashboards cannot mix the new methodology with historical raw-first samples
• assert stable versioned IDs, balanced scheduling, legacy-query ordering, non-mutation, shared metadata, and relation-only SQL differences

Validation

• just test-perf
• just test-scenario
• just lint

Scope and interpretation

The targeted frozen-perf scenario already creates and validates both relation families, so scenario and setup wiring are unchanged. Both targets use the same frozen Parquet objects; this measures view access versus DuckLake metadata/table access, not a production partition-layout rewrite. Alternating order balances which target runs first, but both targets can still be affected by shared caches.

Follow-up

The companion Grafana PR visualizes paired p50/p95 latency, matched-iteration ratios, sample counts, errors, and missing targets using the stable target suffixes and versioned intents.

[github-actions]

Test Impact Plan

Deterministic summary of how this PR changes tests, CI runners, and coverage-risk signals.

Summary

Area	Added	Changed	Deleted
Test files	0	6	0
E2E/journey files	0	0	0
Workflow files	0	0	0

Signals

• Test cases: +4 / -1
• Assertions: +14 / -1
• Skips or known failures added: 0
• Workflow continue-on-error added: 0
• Workflow path filters added: 0
• Test commands removed from justfile: 0
• E2E/journey retry lines added: 0

Coverage risk: neutral or increased

No coverage-reduction warnings detected.