[LLM-Katan] Support Different Reasoning Template

[rootfs]

Is your feature request related to a problem? Please describe.
We have 3 reasoning template in the router config, it would be great for LLM-Katan to support them in e2e test.

# Reasoning family configurations
reasoning_families:
  deepseek:
    type: "chat_template_kwargs"
    parameter: "thinking"

  qwen3:
    type: "chat_template_kwargs"
    parameter: "enable_thinking"

  gpt-oss:
    type: "reasoning_effort"
    parameter: "reasoning_effort"
  gpt:
    type: "reasoning_effort"
    parameter: "reasoning_effort"

[rootfs]

@yossiovadia @JaredforReal @yuluo-yx