Escalate underchallenging hunts from combat pressure

[adrunkhuman]

Summary

• Replace the universal hunt threat ceiling with a bounded per-controller challenge frontier that escalates after sustained easy combat and backs off after recovery pressure, danger, or death.
• Rank rewarding reachable regions near the desired pressure band while rejecting raw predicted-lethal damage against snapshotted current health.
• Predict conservative potion and legal Light Healing recovery from loaded player and spell state, and use the same immutable pressure model for normal and hypothetical equipment evaluation.
• Emit active-combat observations, frontier changes, candidate bands, and stable local-scope exhaustion with delayed reevaluation.

Verification

• pwsh -File scripts/test-playerbot-gameplay.ps1 -AdaptiveChallenge -DeathTelemetry -Focused -SkipBuild -TimeoutSeconds 300 - passed frontier, recovery, lethal, hysteresis, exhaustion, and death scenarios
• pwsh -File scripts/test-playerbot-gameplay.ps1 -HuntRegionPlanning -Focused -SkipBuild -TimeoutSeconds 240 - passed bounded scanning, ranking, reachability, and cancellation scenarios
• pwsh -File scripts/test-playerbot-gameplay.ps1 -FullNavigation -CorpseLoot -Focused -SkipBuild -TimeoutSeconds 300 - passed navigation and corpse-loot regressions

Limits

A one-hour normal-stack observation selected in-band real-map hunts, emitted complete outcomes, and progressed Bot One from level 8 to 9 without death. The five-minute development hunts produced only 6-30% active-combat uptime, so the frontier correctly remained at 0.20 with insufficient_active_combat. Deterministic fixtures prove transition logic; a separate 20-60 minute hunt soak is still required to assess sustained live escalation.

Closes #92.