Re-measure under the honest cost model: simulate re-run, DCA dip ablation, paper DCA re-enabled

[eaitbrahim]

Three measurements that belong together, all enabled by Phase 9 and the 0.8.2 release it needs:

1. Re-run keel simulate under the honest cost model. feat(backtest): per-product slippage scaled from liquidity — assumed, capped, and reported (#259) #334's per-product slippage (5–50bp by liquidity) only reaches the deployment in 0.8.2. The question it answers: do the thin-asset outliers (TON-class gross PF 3.751 on n=9) survive pricing that respects their books, or evaporate? This is the measurement that decides whether Opt the promotion gate into per-product slippage — prerequisite-gated on a thin-asset candidate or a load-bearing restatement #335's promotion-gate opt-in is load-bearing or precautionary.
2. The DCA dip-bonus ablation (proposed by the 2026-08-17 scout run): three cells — dip_bonus_pct 0 / 1 / 2 at constructor defaults — in sim/portfolio_sim (fee_pct=0.012, slippage_pct=0.0005, BTC-USD), reporting terminal sleeve mark-to-market net of entry fees, never a profit factor (the sleeve never closes; DCA is exempt from the promotion gate per §12.6). Pre-registered ordinal expectation from the citation (bestinterest.blog, BTC 2013–2021, costs not modeled): plain cadence ≥ dip-scaled, worse when deeper. First-ever measurement of the only unswept rule family.
3. Re-enable the paper DCA twins (rules 19/20, currently disabled in the paper book) so the paper pipeline stops reading zero signals from the one family that can trade frequently without stacking round-trip tolls — fees are proportional and DCA never exits, so accumulation frequency is toll-free per unit deployed (the status review's asymmetry point).

Acceptance

• 0.8.2 cut and deployed (carries fix(executor): warn loudly when a conditional entry price is overridden by market routing (#260) #332/feat(backtest): per-product slippage scaled from liquidity — assumed, capped, and reported (#259) #334/chore(fidelity): close Phase 9's loose ends — alias, stale docstrings, telemetry overflow guard #336).
• Simulate re-run recorded as a dated experiment document with the per-product slippage table beside the verdict.
• The three-cell ablation run and recorded in the same (or companion) document, instrument and caveat stated.
• Paper DCA twins re-enabled through the supported CLI path (not raw SQL), with the status change noted in the run log.