Fix parallel node functions triggering multiple LLM completions

[omChauhanDev]

Fixes : #243, pipecat-ai/pipecat#3925

When the LLM makes multiple parallel calls to the same node function (common with global_functions), each result was triggering a separate LLM completion — producing multiple responses instead of one.

Root cause: node functions explicitly set run_llm=True on each result, overriding pipecat's built-in logic that already knows to wait until all parallel calls finish.

Fix: run_llm=True → run_llm=None - let pipecat's default handle it.

• Single calls: no behavior change
• Parallel calls: LLM now waits for all results before responding