Accumulates user transcriptions into the LLM context. When the user stops speaking (UserStoppedSpeakingFrame), pushes an LLMContextFrame to trigger LLM processing.

Accepts either a context_agent (shared via ContextAggregatorPair) or a standalone context option.