Support responseSchema in VertexAiGeminiChatOptions

[SuminAndrew]

Closes #2087

VertexAI's GenerationConfig natively uses a Schema object to configure response schema, which represents an OpenAPI schema. There's also a responseJsonSchema option, but it is only supported in Gemini 2.5 (see: https://ai.google.dev/gemini-api/docs/structured-output#json-schema).

To support this, I added a new String responseSchema property to VertexAiGeminiChatOptions. This allows users to provide an OpenAPI schema as a string, which is then converted into a Schema object internally. The same approach is already used for tool definitions via inputSchemas, so I extracted the conversion logic into a shared utility class. I also added corresponding unit tests.

[sobychacko]

@SuminAndrew, could you sign your commit using DCO? https://spring.io/blog/2025/01/06/hello-dco-goodbye-cla-simplifying-contributions-to-spring

[SuminAndrew]

@SuminAndrew, could you sign your commit using DCO? https://spring.io/blog/2025/01/06/hello-dco-goodbye-cla-simplifying-contributions-to-spring

@sobychacko Done

[SuminAndrew]

Hey @sobychacko, I wonder what are the next steps for this PR? Should I tag someone to get this reviewed?

[SuminAndrew]

Updated the branch to fix conflicts with the main branch

[SuminAndrew]

Also tagging @ilayaperumalg
Would be great if this could make it into 1.1.0, this is one of the bits that we are currently missing compared to langchain4j

[SuminAndrew]

Rebased onto a fresh main branch to fix failing MCP Client tests (unrelated to the changeset).

[SuminAndrew]

Another rebase to stay up-to-date

[ilayaperumalg]

@SuminAndrew Sorry about the delay in response. The changes look good. Rebased, verified the tests and build locally, and merged as 31423e9.

[ilayaperumalg]

@SuminAndrew Thank you for adding this feature!