[CI] Restructure model download list by cluster and update workflow (#14908)

### What this PR does / why we need it?

- Restructure model_dataset_list.json from flat array to per-cluster
object (a2/a3/a5), classifying models by their target CI cluster
- Remove ~260 unused models not referenced by any test config
- Add all complete-weight models from A3 (267) and A5 (38) CI disks
- Keep A2 models (34) matching test configurations
- Update labeled_download_model_dataset.yaml to use cluster-aware matrix
and jq extraction, so each runner only downloads its own cluster's
models
- Fix path trigger filename mismatch (labled -> labeled)

### Does this PR introduce _any_ user-facing change?
No

### How was this patch tested?

- vLLM version: v0.27.1
- vLLM main:
https://github.com/vllm-project/vllm/commit/ba07e4a48fc951300d97eb506217dd530583dea3

Signed-off-by: hfadzxy <starmoon_zhang@163.com>