[BUG] Klat does not handle vLLM deployment updates

[NeonDaniel]

Description

After restarting vLLM, Klat continues to request the old models rather than the updated ones.

Steps to Reproduce

• Update a vLLM deployment
• Restart the neon-llm-vllm container
• Request one of the vLLM chatbots in Klat
• Observe errors logged in neon-llm-vllm container (also reported to Sentry)

Relevant Code

No response

Other Notes

Previously, these invalid requests would fail; they now log errors with NeonGeckoCom/neon-llm-vllm#9

[NeonDaniel]

@NeonKirill if you can point me at the place(s) where Klat receives available models from the vLLM container, I can probably troubleshoot and resolve this.