Use Qwen 3.6 27B with rate-aware cloud queueing

[pgotta]

What changed

• pins Groq cloud inference to qwen/qwen3.6-27b
• migrates stale Scout model preferences automatically
• resizes summary, detailed, chapter, and Ask requests to fit beneath Qwen's free-tier 8K TPM limit
• queues calls against conservative TPM/RPM/daily budgets
• follows Groq's live remaining-token and reset headers
• distinguishes minute limits from organization-wide daily exhaustion
• keeps existing checkpoints/resume behavior
• updates Settings, versioning, README, and changelog for 1.4.0

API keys

This intentionally keeps one API-key slot. Groq quotas are organization-wide, so extra keys in the same organization do not add capacity. Automatic rotation after quota exhaustion would also bypass the provider's intended limits. For more daily capacity, use Groq's Developer tier.

Verification

The Android GitHub Actions build should run on this PR.