fix(llm): prevent base llm flow truncation (#522)

[AlankritVerma01]

Fixes: #522

Previously, BaseLlmFlow.run_async would loop indefinitely whenever the LLM response was truncated due to hitting its max_tokens limit. The loop’s break condition only checked for “no events” or last_event.is_final_response(), neither of which fires for a partial/truncated chunk—so the generator re-invoked _run_one_step_async endlessly.

With this change, we:

• Treat any partial (truncated) event as terminal by adding or last_event.partial to the break condition in run_async.
• Ensure that when an LLM emits a truncated chunk (partial=True), run_async yields it exactly once and then terminates, avoiding infinite loops.

Test Plan

• New unit test test_base_llm_flow_truncation.py verifies that:
  • Given a flow that only ever yields one partial event, run_async emits exactly one event and then stops.
• Run it with:

  pytest tests/unittests/flows/llm_flows/test_base_llm_flow_truncation.py

[AlankritVerma01]

Hey @Jacksunwei @hangfei
Please review the PR
Thanks team!