feat(groq): native content-block streaming events

[LennyMalcolm0]

Stacked on #14 (core BlockStreamTracker + Anthropic), which it branches from. Depends only on the core primitive — sibling of the OpenAI chain. Retargets to master once #14 lands.

Gives ChatGroq native _stream_chat_model_events / _astream_chat_model_events hooks so stream_events(version="v3") maps groq's stream directly to content-block events.

How it works

A bespoke converter (convert_groq_stream / async twin) builds text, reasoning, and tool-call blocks directly from groq's raw OpenAI-shaped delta, feeding the shared BlockStreamTracker. groq streams tool-call arguments incrementally (id+name on the first delta, fragments on later deltas at the same index), so the tracker accumulates them and finalizes to a parsed dict. Usage is read from x_groq.usage; the arguments == "null" quirk is normalized to "{}".

It deliberately does not depend on langchain-openai — it reuses only groq's own _create_usage_metadata.

The win over the bridge

groq's existing chunk parser puts tool_calls and reasoning in additional_kwargs, so the compat bridge surfaces neither as content blocks. The native converter surfaces both — tool calls as tool_call blocks and reasoning as reasoning blocks.

Worth careful review

• Tool-arg accumulation across deltas (keyed by tool:{index}); a test covers parallel tool calls at distinct indices.
• message_id is keyword-only on the hooks, kept out of the request kwargs; message-start carries core's LangChain run id.
• Scope cut: groq server-tool results (executed_tools) are not yet surfaced as blocks (noted in-code); the bridge still carried them in additional_kwargs.
• Sync/async are faithful twins.

---

Cloned from langchain-ai#38029 for DevAsign stress testing.