feat: implement JSON-driven conversational flow engine

Replaces hardcoded ConversationHandlers with a generic flow engine that reads conversation definitions from talia_bot/data/flows.json.

- Adds a 'conversations' table to the database to persist user state, making flows robust against restarts.
- Implements a central 'universal_handler' in main.py to process all user inputs (text, voice, callbacks, documents) through the new engine.
- Refactors Vikunja, LLM, and Calendar modules to be asynchronous and support the new architecture.
- Adds a new 'transcription' module for OpenAI Whisper and a 'mailer' module for the print flow.
- Implements the full logic for all specified user flows, including project/task management, calendar blocking, idea capture (with branching logic), and the RAG-based client sales funnel.
- Cleans up legacy code and handlers.

talia_bot/db.py

@@ -32,8 +32,20 @@ def setup_database():
             )
         """)
 
+        cursor.execute("""
+            CREATE TABLE IF NOT EXISTS conversations (
+                id INTEGER PRIMARY KEY AUTOINCREMENT,
+                user_id INTEGER NOT NULL,
+                flow_id TEXT NOT NULL,
+                current_step_id INTEGER NOT NULL,
+                collected_data TEXT,
+                updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+                FOREIGN KEY (user_id) REFERENCES users (telegram_id)
+            )
+        """)
+
         conn.commit()
-        logger.info("Database setup complete. 'users' table is ready.")
+        logger.info("Database setup complete. 'users' and 'conversations' tables are ready.")
     except sqlite3.Error as e:
         logger.error(f"Database error during setup: {e}")
     finally: