AI Automation Suggester 1.5.7

This patch fixes a setup failure on recent Home Assistant releases and a Perplexity setup-validation error.

Highlights

• The integration no longer runs an AI inference during sensor platform setup, which previously caused setup_error on Home Assistant 2025.x and newer.
• Perplexity setup validation now succeeds with sonar models.

Fixed

• Setup error on HA 2025.x+ (issue #166): Sensor setup passed update_before_add=True, which in newer Home Assistant versions calls CoordinatorEntity.async_update() during platform setup. That triggered a full LLM request (10–60+ seconds) and exceeded the setup timeout, leaving the integration in setup_error on every restart — most visibly with local/slow providers (Ollama, LocalAI, llama.cpp). Entities now register without an immediate refresh and populate from coordinator data, refreshing when generate_suggestions runs.
• Perplexity validation always failing (issue #171): The setup validation probe sent max_tokens: 1, but Perplexity sonar models require max_tokens >= 16 and reject smaller values with a 400. The probe now uses max_tokens: 16, so validation passes regardless of the user's configured output token limit.

Upgrade notes

No configuration changes are required. After updating, restart Home Assistant; the integration should load without setup_error, and Perplexity setup validation should succeed.