feat(nlp): add offline Smart Natural Language Commands dashboard

[MatinHoseiny]

Description

This PR introduces a 100% offline, local natural language command dashboard that allows users to manage local music collections using plain English text commands.

Key features:

• Zero API / Network Dependencies: The command parser is lightweight, offline, and runs purely on local regexes and background coroutines.
• Conjunction Splitting: Parses commands containing multiple artists or genres seamlessly (e.g. "create playlist of Pop Smoke and Queen").
• Typo Tolerance: Implements a space-optimized Levenshtein distance algorithm to match queries with database entries even with spelling mistakes.
• Direct Integration: The Smart screen is added as the fourth tab in the bottom navigation bar with a matching theme-friendly icon.
• Safety Dialogs: Destructive actions (artist bulk deletion) are protected by a secure 2-step confirmation dialog showing the exact local file paths to be deleted.

Files Added:

• data/nlp/: Contains the intents, command parser, fuzzy matching engine, repository, and result models.
• presentation/nlp/: Contains the Compose screen dashboard and its state ViewModel.
• NlpCommandParserTest.kt: A comprehensive suite of unit tests validating parsing, splitting, and matching offline.