Add runInParallel option to input guardrail initialization

[seratch]

The Python SDK now has a new option "run_in_parallel", which enables agent apps to hold off LLM and tool calls until input guardrail executions complete. This TS SDK will have the same in future verisons.

References:

• feat: add run_in_parallel parameter to input guardrails openai-agents-python#1986