fix: duplicate background tasks

[Akinator31]

This pull request introduces several improvements and fixes across the codebase, focusing on database reliability, concurrency, error handling, and code modernization. The most significant changes include making thread closure and message number allocation atomic and robust, improving error logging, updating user fetching methods for consistency, and adding comprehensive tests for database operations.

Database reliability and concurrency improvements:

• Refactored close_thread to be idempotent, returning a boolean to indicate if the thread was actually closed, and ensuring concurrent close attempts do not overwrite each other's metadata. The SQL update now only affects open threads (status = 1). [1] [2]
• Rewrote allocate_next_message_number to use a single atomic SQL statement with RETURNING, ensuring unique message numbers even under heavy concurrency, and added a test to verify uniqueness. [1] [2]
• Reduced SQLite connection pool size and set a shorter acquire timeout to reduce lock contention and fail fast under load.

Error handling and logging:

• Replaced silent ignores of database update errors (e.g., in update_thread_status_db) with explicit error logging using eprintln! in multiple commands, improving observability of failures. [1] [2] [3] [4] [5] [6] [7]

User fetching modernization:

• Updated all usages of ctx.http.get_user to use the more idiomatic UserId::to_user(&ctx) method, ensuring consistency and leveraging the latest library patterns. [1] [2] [3] [4] [5]

Testing and code quality:

• Added comprehensive tests for thread closing and message number allocation, including concurrency checks, to the update_thread_status_db module, improving code safety and reliability.

Other codebase improvements:

• Introduced a global keyed task manager for reminder timers using LazyLock, ensuring one timer per reminder ID and improving reminder scheduling. [1] [2] [3]
• Added utility types (FlagGuard) for managing atomic flags in the ready handler, ensuring proper cleanup even if tasks panic.

These changes collectively improve the robustness, maintainability, and observability of the application.