feat: support llm filter for rss watcher

[fanqiNO1]

No description provided.

[Gennadiyev]

If I were you I would use response_format throughout... But yours have compatability so that's a reasonable trade-off.

[Copilot]

Pull request overview

This PR adds LLM-based filtering capability to the RSS watcher, allowing users to automatically filter RSS feed entries based on relevance before processing them for digests or notifications. This helps reduce noise from high-volume feeds by only showing entries deemed relevant by an LLM.

Changes:

• Added LLM filter configuration options with customizable prompts, model selection, and concurrency limits
• Implemented filter_entries_with_llm method that scores each entry on a 0-5 scale and filters based on threshold
• Updated logging levels for cleaner output (changed some log.warning to log.debug, removed verbose debug messages)

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 10 comments.

File	Description
watchers/rss_watcher.py	Added LLM filter implementation with new configuration parsing, async filtering logic using semaphores for concurrency control, and integrated filtering into the entry processing pipeline
config.example.yaml	Added example configuration for LLM filter feature showing model selection, custom prompt, and concurrency settings