feat(qwen): add Qwen3.8-27B fine-tuning recipes (#3560)

* feat(qwen): add Qwen3.8-27B fine-tuning recipes

Signed-off-by: Alexandros Koumparoulis <akoumparouli@nvidia.com>

* ci(vlm): declare recipe_owner and nodes on qwen3.8-27b recipes

Both new recipes are auto-enrolled in CI by their presence under
examples/vlm_finetune/, so validate_new_recipe_ci.py requires a complete
ci: block. They declared only time and env_vars.

Both are pure FSDP2 (tp/cp/pp/ep all 1) and documented as
--nproc-per-node 8, so nodes: 1.

Signed-off-by: Alexandros Koumparoulis <akoumparouli@nvidia.com>

* refactor(vlm): move qwen3.8-27b recipes to examples/vlm_finetune/qwen3_8

The recipes target Qwen/Qwen3.8-27B but sat under the qwen3_5 folder.
generate_ci_tests derives MODEL_FAMILY from the recipe's parent directory,
so they were reported as qwen3_5 in CI.

Updates the self-referencing header comments, README, model-coverage and
guide docs, and the recipe-contract test's RECIPE_DIR.

Signed-off-by: Alexandros Koumparoulis <akoumparouli@nvidia.com>

---------

Signed-off-by: Alexandros Koumparoulis <akoumparouli@nvidia.com>