refactor: consolidate LLM JSON parsing into shared llm_utils module

Extracted duplicate _strip_markdown_fences / _parse_llm_json functions
from 7 files into app/services/llm_utils.py. Two shared functions:
- strip_markdown_fences(): fence stripping only
- parse_llm_json(): fence stripping + JSON parse + error logging

Files updated: flowpilot_engine, knowledge_flywheel, session_to_flow_service,
ai_tree_generator_service, ai_fix_service, ai_chat_service, kb_conversion_service

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

backend/app/services/llm_utils.py

@@ -0,0 +1,39 @@
+"""Shared utilities for parsing LLM responses."""
+
+import json
+import logging
+from typing import Any
+
+logger = logging.getLogger(__name__)
+
+
+def strip_markdown_fences(text: str) -> str:
+    """Strip markdown code fences from LLM output, returning raw content.
+
+    Use this when you need just the stripping without JSON parsing
+    (e.g., when the caller has its own error handling for json.loads).
+    """
+    text = text.strip()
+    if text.startswith("```"):
+        lines = text.split("\n")
+        lines = [line for line in lines if not line.strip().startswith("```")]
+        text = "\n".join(lines).strip()
+    return text
+
+
+def parse_llm_json(raw_text: str) -> dict[str, Any]:
+    """Parse JSON from LLM response, handling common quirks.
+
+    Strips markdown code fences (```json ... ``` or ``` ... ```) if present,
+    then parses the remaining text as JSON.
+
+    Raises:
+        ValueError: If the text is not valid JSON after fence stripping.
+    """
+    text = strip_markdown_fences(raw_text)
+
+    try:
+        return json.loads(text)
+    except json.JSONDecodeError as e:
+        logger.warning("LLM JSON parse failed: %s — raw: %.300s", e, text)
+        raise ValueError(f"Invalid JSON from LLM: {e}") from e