fix: groq service_tier is OnDemand

[TwistingTwists]

got async-openai deserialization errors for groq on free tier.

ERROR async_openai::error: failed deserialization of: {"id":"chat04-db87f6e7f5e7","object":"chat.comple
   tion","created":1761988004,"model":"moonshotai/kimi-k2-instruct-0905","choices":[{"index":0,"message":{"role":"assistant","content":"```mermaid   K\n```"},"logprobs":null,"finish_reason":"stop"}],"usage":{"queue_time":0.243496693,"prompt_tokens":368,"prompt_time":0.046633806,"completion_tokens":268,"completion_time":0.492293986,"total_tokens":636,"total_time":0.538927792},"usage_breakdown":null,"system_fingerprint":"fp","x_groq":{"id":"req_zapj1rejb9pdn"},"service_tier":"on_demand"}

[TwistingTwists]

closing this in favour of : https://github.com/64bit/async-openai/tree/main/examples/bring-your-own-type
byot is a feature.