feat(js/plugins/google-genai): Support for lyria 3 in vertexAI

[ifielker]

Also some quality of life improvements (that kind of snowballed, sorry):

• response modalities for image models and audio models now have defaults
• error message details are now included in the error. (and formatted nicely)
• nano banana 2 image sizes are updated

Checklist (if applicable):

• PR title is following https://www.conventionalcommits.org/en/v1.0.0/
• Tested (manually, unit tested, etc.)
• Docs updated (updated docs or a docs bug required)

[gemini-code-assist]

Code Review

This pull request adds support for new Gemini and Lyria model features, including response modalities, thinking configurations, and the Vertex AI interactions endpoint. Key updates include automated defaulting of response modalities for TTS and image models, enhanced error detail parsing in both Google AI and Vertex AI clients, and the addition of Lyria 3 model support. The review feedback focuses on improving error handling consistency by using GenkitError with specific statuses, refining imports, and removing unnecessary array copies. Additionally, it is recommended to align error message formatting across different client implementations to ensure a uniform user experience.