Assorted updates, primarily to support Gemini 2.0 (fixes #7345)

• Extensive testing across multiple models
• Extensive testing across both the GCP and Google AI Studio platforms
• Backward and (limited) forward compatibility for grounding with Google Search tool (related to stuff found with PR feat(google-common): Grounding with Google Search and Vertex AI Search #7280)
• Add grounding detailed results to response (related to google-common [Feature]: Grounding with Google Search and Vertex AI Search #5073 and feat(google-common): Grounding with Google Search and Vertex AI Search #7280)
• Do not use API Key for Google Cloud Platform (fixes issues google-vertexai model authentication error - throw new Error("APIs that require a project ID cannot use an API key") #5000 and GOOGLE_API_KEY and GOOGLE_APPLICATION_CREDENTIALS conflict in .env #7399)

Some things to note:

• The tests against "gemini-2.0-flash-thinking-exp" have been commented out. While it works for some things, this model doesn't support tools yet, and we need to look at how to handle the reasoning tokens which are provided. (See google-common: Support Gemini 2.0 Thinking model #7434)
• Functions work with Gemini 2.0, however the prompting for them is more finicky, at least with the configuration used in the testing. (See https://issuetracker.google.com/issues/384752742 where I've opened the issue with Google.) But the functionality itself works.
• This does not account for multimodal outputs, which are expected "soon". Can't do anything with that till we get official info on how it will work.