feat: complete phase 2 -- multi-agent routing, interrupt TTL, escalation, templates

- Intent classification with LLM structured output (single/multi/ambiguous)
- Discount agent with apply_discount and generate_coupon tools
- Interrupt manager with 30-min TTL auto-expiration and retry prompts
- Webhook escalation module with exponential backoff retry (max 3)
- Three vertical industry templates (e-commerce, SaaS, fintech)
- Template loading in AgentRegistry
- Enhanced supervisor prompt with dynamic agent descriptions
- 153 tests passing, 90.18% coverage

backend/templates/e-commerce.yaml

@@ -0,0 +1,42 @@
+agents:
+  - name: order_lookup
+    description: "Looks up order status and tracking information. Use for queries about order status, shipping, and delivery."
+    permission: read
+    personality:
+      tone: "friendly and informative"
+      greeting: "I can help you check your order status!"
+      escalation_message: "Let me connect you with our support team for more details."
+    tools:
+      - get_order_status
+      - get_tracking_info
+
+  - name: order_actions
+    description: "Performs order modifications like cancellations. Use when the customer wants to cancel, modify, or change an order."
+    permission: write
+    personality:
+      tone: "careful and reassuring"
+      greeting: "I can help you with order changes."
+      escalation_message: "I'll connect you with a specialist who can assist further."
+    tools:
+      - cancel_order
+
+  - name: discount
+    description: "Applies discounts and generates coupon codes. Use when the customer asks for discounts, promotions, or coupon codes."
+    permission: write
+    personality:
+      tone: "generous and accommodating"
+      greeting: "I can help you with discounts and coupons!"
+      escalation_message: "Let me connect you with our promotions team."
+    tools:
+      - apply_discount
+      - generate_coupon
+
+  - name: fallback
+    description: "Handles general questions, unclear requests, and conversations that don't match other agents."
+    permission: read
+    personality:
+      tone: "professional and helpful"
+      greeting: "Hello! How can I help you today?"
+      escalation_message: "Let me connect you with a human agent who can better assist you."
+    tools:
+      - fallback_respond