[Docs] Update examples

[peterschmidt85]

• Update Services, SGLang, vLLM, NIM, and TensorRT-LLM examples (migrate to the new models)
• Update TensorRT-LLM example (replace the legacy DeepSeek/Triton flow with an inline Qwen3 example)
• Remove Models section with Wan2.2 (logic here is its' better to have up-to-date Inference examples than out-dated and incomplete Models section)
• Remove deprecated example, incl. Llama 3.1, Llama 3.2, Deepseek, and other examples that we previously unlisted
• Rely on "inline" examples only (logic here is to keep the minimum of maintainable examples, and do not aim for full reproducibility - this requires too much effort to maintain and gets outdated rather quickly) - WIP
• Get rid out of "examples generation"; move all examples from /examples to docs and remove the corresponding hook (logic here is to do not complicate examples - see above)
• MAJOR Update TRL examples under Single-node training
• MAJOR Update AMD example
• MAJOR Update TRL examples under Distributed training
• Update TPU example

[peterschmidt85]

@Bihan please review the complete changes; if all is correct, I'll merge it and move the rest of the mentioned items above into a separate issue/PR. Plus share any feed back on the not-complete items above if you agree/disagree.