Update TRL example paths after examples/ reorg

[sergiopaniego]

Follow-up to huggingface/trl#6820: TRL's examples/ was reorganized into self-contained per-example folders, so links to the old examples/scripts/... paths will 404 once it merges. This updates them to the new locations.

[sergiopaniego]

Extended to the older posts too: sft.py links now point to the maintained trl/scripts/sft.py, renamed VLM scripts (vsft_llava.py, dpo_visual.py) to their current successors, and removed scripts (ddpo.py, gold.py) are pinned to the last release tag that contains them. delta-weight-sync.md needs no changes (it installs from the frozen delta-weight-sync branch). cc @pcuenca

[pcuenca]

Thanks for keeping them up to date!

[sergiopaniego]

Note for merging: this should land after huggingface/trl#6820 does (the new paths don't exist on TRL main until then).

[qgallouedec]

Thanks for the sweep! One catch: huggingface/trl#6820 also renames each example's main file after its folder, so 6 of the new links still 404 on the reorg branch:

• examples/dpo_reduce_hallucinations/dpo_vlm.py -> examples/dpo_reduce_hallucinations/dpo_reduce_hallucinations.py (dpo_vlm.md, zh/dpo_vlm.md, smolvlm.md)
• examples/sft_visual_chat/sft_vlm.py -> examples/sft_visual_chat/sft_visual_chat.py (llama32.md, zh/llama32.md, vlms.md, zh/vlms.md, trl-vlm-alignment.md)
• examples/grpo_visual_math/grpo_vlm.py -> examples/grpo_visual_math/grpo_visual_math.py (trl-vlm-alignment.md)
• examples/rloo_visual_math/rloo_vlm.py -> examples/rloo_visual_math/rloo_visual_math.py (trl-vlm-alignment.md)
• examples/online_dpo_visual_math/online_dpo_vlm.py -> examples/online_dpo_visual_math/online_dpo_visual_math.py (trl-vlm-alignment.md)
• examples/async_grpo_opencode/opencode.py -> examples/async_grpo_opencode/async_grpo_opencode.py (muse-glimmer.md)

The rest checks out, I validated every new link against the reorg-examples branch (the trl/scripts/sft.py retargets and the v1.10.0-pinned ones all resolve).