feat(vertex_ai): add Gemini 3.5 Flash model support

[fuyundev]

Summary

Adds the gemini-3.5-flash model to the vertex_ai plugin.

• New model YAML: models/vertex_ai/models/llm/gemini-3.5-flash.yaml — mirrors the parameter style of the existing Gemini 3 family (thinking_level, include_thoughts, media_resolution, grounding, url_context, code_execution, json_schema)
• Registers the model in GLOBAL_ONLY_MODELS_DEFAULT in llm.py
• Bumps plugin version 0.0.54 → 0.0.55 (0.0.54 was consumed by fix(vertex_ai): build typed Gemini tool schemas #3178)

Notes on prior review feedback (#3167)

The previous PR #3167 received automated review comments that I verified against the codebase and official docs — leaving them as-is for the following reasons:

1. Pricing format (string vs numeric): 773 model YAMLs in this repo use quoted strings; 0 use bare numerics. The recently-merged gemini/gemini-3.5-flash.yaml (feat(gemini): add Gemini 3.5 Flash model support and refactor service_tier logic #3165) and aihubmix/gemini-3.5-flash.yaml (feat(aihubmix): add Qwen3.7 Max, Gemini 3.5 Flash, Grok 4.3 models and improve Anthropic support #3179) both use string pricing. Switching this one file to numeric would make it the only inconsistent file.

2. max_output_tokens: 65536: Matches Google's official spec for Gemini 3.5 Flash (output limit = 65,536) and the rest of the Gemini 3 family in this repo.

3. Pricing $1.5 / $9.0 per 1M tokens: Official Gemini 3.5 Flash pricing announced 2026-05-19. Same values as the merged gemini and aihubmix YAMLs. The comparison to Gemini 2.0 Flash isn't apples-to-apples — 3.5 Flash is a different generation and capability tier.

Refs:

• https://ai.google.dev/gemini-api/docs/models/gemini-3.5-flash
• https://ai.google.dev/gemini-api/docs/pricing

Test plan

• Verify model appears in Dify model picker for vertex_ai
• Smoke-test a chat completion with thinking_level=Medium
• Verify grounding and url_context toggles work

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request adds support for the Gemini 3.5 Flash model to the Vertex AI provider. The changes include a new YAML configuration file defining the model's features, parameters, and pricing, as well as the registration of the model in the Python codebase and a version bump in the manifest. I have no feedback to provide as there were no review comments.