seats-aero: add layover filter and mandate trip-detail drilling

[BarnNorth]

First off — this toolkit is excellent. I used it to research and book a transatlantic business class honeymoon flight (LAX→ZRH on the Lufthansa A380, 90K Aeroplan via Chase UR transfer) and it genuinely made a complex, high-stakes booking feel manageable. The seats.aero workflow, transfer partner optimization, and award-holds guidance all worked exactly as intended. This change comes from one rough edge I hit during that session.

Problem

Searching for LAX→ZRH business class on Aeroplan, the skill surfaced the cheapest result: 70K Saver on LH457+LX1075 with an 8-hour layover in Frankfurt. It presented this without comment. The actual best option was 90K Standard on LH453+LX1105 — the Lufthansa A380 with a 1h35m layover in Munich. 20K more miles, 6.5 fewer hours sitting in an airport.

Two issues caused this:

1. The search summary object (JMileageCost) only reports the cheapest itinerary across all trips. Trip details were never pulled, so the better Standard-tier routing was never surfaced.
2. Even if trip details had been pulled, there was no filter to flag the 8-hour layover as worth reconsidering.

Changes

• Make GET /trips/{id} a mandatory step in the workflow for any promising result — the search summary shows cheapest only, trips show everything
• Add a Layover Quality Filter section: flag any layover over 3 hours, automatically surface the next-best itinerary with a shorter layover alongside it, let the user decide whether the miles savings are worth the wait

Before / After

Before: Skill returns JMileageCost: 70000 from the search summary and stops. User sees one option: LH457+LX1075, 70K, 8h FRA layover. No flag, no alternatives shown.

After: Skill pulls trip details, calculates layover per itinerary, and presents:

Flight	Miles	Layover	Flag
LH457+LX1075	70,000	8h FRA	⚠️ layover over 3h
LH453+LX1105	90,000	1h35m MUC	✓ next-best alternative

Test plan

• bash scripts/smoke-test.sh --quick — 0 new failures (2 pre-existing on upstream main: stale data files + macOS missing timeout)
• All 44 skills have valid frontmatter
• README and llms.txt unchanged (no frontmatter edits)