feat(engine): replace TinkerBackendServer with in-process local handler

[luyuzhe111]

Summary

This PR simplifies the integration or Tinker with rllm-model-gateway by replacing the TinkerBackendServer (a standalone http server) with a simple local handler; also adds tool call support.

Type of change

• Feature
• Fix
• Docs
• Refactor
• Example / Project
• Infra / CI

What changed

• Inject an async handler directly into the gateway's ReverseProxy instead of running a sidecar HTTP server for TinkerEngine.
• Add tool_calls parsing to trace_converter and integration tests for the full tinker → gateway → trace → step pipeline.
• Updated tests/conftest.py to try importing first so that integration can run when the dependencies are available.

Validation

• pre-commit run --all-files
• Targeted tests: pytest ...
• Manual validation performed
• Not run (reason below)

Validation details:

• unit test: pytest tests/engine/test_trace_converter.py
• integration test: pytest tests/integration/test_tinker_gateway_integration.py (TINKER_API_KEY env var required).

Breaking changes / migration notes

• None

Docs / examples

• Not needed
• Updated docs
• Updated examples
• Follow-up docs needed

Related issues / PRs

• Fixes #
• Related to #
• Stacked on / depends on #

Screenshots / logs