fix: tool registry emits Anthropic schema, question lifecycle cleanup

- Tool.to_schema() emits input_schema (Anthropic Messages API) instead
  of OpenAI-style parameters
- QuestionManager: remove questions from _pending on answer and timeout,
  reject double answers, add list_pending()
- core/skills/__init__.py so skills are importable as a package
- revert optional-import fallbacks in logger/message_bus: nats-py and
  loguru are hard requirements, a broken install should fail loudly
- tests for wait_for_answer happy path and timeout cleanup

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012AJLYcbTHCe45fyFnMfEin

core/message_bus.py

@@ -8,11 +8,8 @@ from collections import defaultdict
 from dataclasses import dataclass
 from enum import Enum
 
-try:
-    import nats
-    import nats.js.errors
-except ImportError:
-    nats = None
+import nats
+import nats.js.errors
 
 from core.types import AgentMessage, MessageType
 from core.logger import logger
@@ -59,9 +56,6 @@ class MessageBus:
 
     async def connect(self):
         """Connect to NATS and create the logistics JetStream stream."""
-        if nats is None:
-            logger.warning("nats-py not installed; message bus running in local in-memory mode")
-            return
         self._nc = await nats.connect(
             servers=[f"nats://{NATS_HOST}:{NATS_PORT}"],
             user=NATS_USER,