feat: add --dry-run VRAM/size estimation mode

[mvanhorn]

Re-submit of #1592, reworked per your feedback. The VRAM estimate is now built on auto-round's own block-wise memory model instead of a raw parameter count:

• Reuses auto_round.utils.device.estimate_tuning_block_mem and get_moe_memory_ratio (@xin3he's card_0 = block_input_output + layer_activation + additional formula).
• Models peak memory at decoder-block granularity (@wenhuach21).
• Excludes the block input/output cache when low_gpu_mem_usage is set, and includes it otherwise (@wenhuach21's caching point).
• MoE handled via get_moe_memory_ratio; layer/hidden-size discovery is robust across config field names and nested configs (text_config etc.) since num_hidden_layers doesn't cover every model (@xin3he).
• Loads AutoConfig only, no weights.

Unit tests cover helper reuse, the low_gpu_mem_usage path, MoE, and layer-count fallbacks (8 passing). Supersedes #1592 (couldn't reopen after the rebase).

[mvanhorn]

Pushed a fix for the two red checks:

• Code-Scan (pylint E0401): the optional modelscope import in _load_auto_config now carries # pylint: disable=E0401, matching how calib_dataset.py and utils/model.py handle modelscope (it's not in requirements). Scan is now 10.00/10 locally.
• DCO: signed off all commits on the branch.

[wenhuach21]

Hi, thanks for the PR! It would be very useful if the estimation is both general and accurate.

Could you kindly share some evaluation results demonstrating its accuracy? It would be helpful to include results covering an LLM, a VLM, and an MoE model.

[mvanhorn]

Happy to. Since --dry-run estimates VRAM/size analytically from the model config rather than measuring on hardware, I can put together a table comparing the estimate against the theoretical footprint (param count x bit-width, plus the group/zp/scale overhead the estimator accounts for) for a representative LLM, VLM, and MoE. That validates the estimation math end to end without needing a specific GPU.

Where I'd lean on you: confirming the estimate against actual measured peak VRAM on real hardware for those three model classes - you're better set up for that than I am. If a config-vs-theoretical table plus your measured spot-check works, I'll post the table. Which three models would you want represented (so the numbers match something you can cross-check)?