[Refactor] Simplify dummy data generation

[DarkLight1337]

Purpose

• Always pass the kwargs from MM config when constructing HF processor. This removes the need to pass them explicitly to dummy data generation ([Multimodal] Expose mm_processor_kwargs for DummyInputsBuilder #34330)
• Always pass a dictionary to mm_options, so there is no need for implementations to handle the case where it is None

Test Plan

Using the same test as #34330

Test Result

(Worker_TP2 pid=1233399) INFO 02-21 15:00:08 [gpu_model_runner.py:5150] Encoder cache will be initialized with a budget of 114688 tokens, and profiled with 1 video items of the maximum feature size.
(Worker_TP3 pid=1233400) INFO 02-21 15:00:08 [gpu_model_runner.py:5150] Encoder cache will be initialized with a budget of 114688 tokens, and profiled with 1 video items of the maximum feature size.
(Worker_TP1 pid=1233398) INFO 02-21 15:00:08 [gpu_model_runner.py:5150] Encoder cache will be initialized with a budget of 114688 tokens, and profiled with 1 video items of the maximum feature size.
(Worker_TP0 pid=1233397) INFO 02-21 15:00:08 [gpu_model_runner.py:5150] Encoder cache will be initialized with a budget of 114688 tokens, and profiled with 1 video items of the maximum feature size.

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[dosubot]

Related Documentation

Checked 0 published document(s) in 1 knowledge base(s). No updates required.

^{How did I do? Any feedback?}  

[mergify]

Documentation preview: https://vllm--35025.org.readthedocs.build/en/35025/

[gemini-code-assist]

Code Review

This pull request refactors the dummy data generation for multimodal models by centralizing the merging of keyword arguments from the multimodal configuration. This simplifies the get_dummy_mm_data and get_hf_processor methods across numerous model implementations. While the refactor is comprehensive, there are a few inconsistencies in the updated method signatures and missing keyword argument support in some custom processor adapters that could lead to runtime errors.