Add ModerationAdvisor for ChatClient

Expected Behavior

It would be helpful to have a built-in ModerationAdvisor that integrates ModerationModel into the ChatClient advisor chain for both non-streaming and streaming flows.

• It evaluates user input before model invocation.
• If moderation is flagged, it returns a configured fallback response.
• If not flagged, it continues to the next advisor.
• It is configurable with builder options such as order, failureResponse, and failure policy (e.g. failOpen).

Example:

ModerationAdvisor moderationAdvisor = ModerationAdvisor.builder()
        .moderationModel(moderationModel)
        .failureResponse("I'm unable to respond to that due to safety policy.")
        .order(0)
        .failOpen(true)
        .build();

ChatClientResponse response = chatClient.prompt()
        .advisors(moderationAdvisor)
        .user(userInput)
        .call()
        .chatClientResponse();

Current Behavior

There is currently no built-in advisor that gates ChatClient requests via ModerationModel.

• SafeGuardAdvisor exists, but it is rule-based/keyword-based.
• For model-based moderation, users currently need to manually call moderation APIs and wire the logic in both call and stream paths.
• This is workable, but tends to add repeated application-level code.

Context

A built-in ModerationAdvisor would provide a reusable way to apply moderation in the advisor chain.