Add tool governance advisors and reasoning trace observability

[vaquarkhan]

Summary

This PR adds a set of standalone ChatClient advisors for tool governance and loop observability in spring-ai-agent-utils:

• ToolCallPolicyAdvisor for allow/deny/rewrite decisions before tool execution.
• ToolApprovalAdvisor for human-in-the-loop approvals using CompletableFuture blocking.
• ToolArgumentValidationAdvisor for tool argument byte-size limits.
• ToolTimeoutAdvisor for per-tool timeout enforcement (with default + overrides).
• ToolRateLimitAdvisor for per-request (loop) tool call rate limiting.
• ToolFilterAdvisor for include/exclude filtering by glob-style tool name patterns.
• ReasoningTraceAdvisor for per-iteration trace emission with timing and tool-call metadata.

Additional changes:

• Added focused unit tests for all new advisors.
• Added docs pages for each advisor under spring-ai-agent-utils/docs.
• Updated spring-ai-agent-utils/README.md advisor section to include all new advisors.

Why

These advisors provide missing governance and observability controls commonly needed in production agent systems:

• enforce policy and approval on tool usage,
• bound unsafe/expensive tool calls (size, timeout, rate),
• scope tool exposure by pattern,
• improve loop-level traceability for operations and debugging.

Test Plan

• Ran advisor-focused unit tests:
  • ToolCallPolicyAdvisorTest
  • ToolApprovalAdvisorTest
  • ToolArgumentValidationAdvisorTest
  • ToolTimeoutAdvisorTest
  • ToolRateLimitAdvisorTest
  • ToolFilterAdvisorTest
  • ReasoningTraceAdvisorTest
    -- Result: all tests passed (15 tests, 0 failures).

Command used:

mvn -f spring-ai-agent-utils/pom.xml "-Dtest=ToolCallPolicyAdvisorTest,ToolApprovalAdvisorTest,ToolArgumentValidationAdvisorTest,ToolTimeoutAdvisorTest,ToolRateLimitAdvisorTest,ToolFilterAdvisorTest,ReasoningTraceAdvisorTest" test