Fix infinite loop in processing graph conditional routing

[Copilot]

The processing graph had conflicting edge definitions that could cause infinite loops when tools are unavailable or fail. The conditional routing from llm node was overridden by an unconditional edge, and there was no protection against repeated tool invocations.

Graph structure fix

• Removed conflicting unconditional llm → END edge that overrode conditional routing logic
• Conditional edges now properly control flow: llm → tools (if requested) or llm → END

Loop prevention

• Added MAX_TOOL_ITERATIONS = 10 constant as recursion limit in graph compilation
• Prevents unbounded llm → tools → llm cycles

Tool error handling

• tool_node now handles missing tools gracefully via .get() instead of direct dict access
• Returns error ToolMessages instead of raising KeyError
• Catches and reports tool execution exceptions

# Before: Would crash if tool unavailable
tool_inst = TOOLS[call["name"]]  # KeyError if tool commented out

# After: Returns error message to LLM
tool_inst = TOOLS.get(call["name"])
if tool_inst is None:
    content = f"Error: Tool '{call['name']}' is not available."

These changes allow the graph to terminate gracefully even when processing-specific tools are disabled in tools/__init__.py.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[iamtekson]

looks good. Please fix my comment on this PR.

[iamtekson]

move this block to the config -> settings.py file