feat: Introduce ManagedAgent and AgentRunner implementations

[jsonbailey]

feat: Add OpenAIAgentRunner with agentic tool-calling loop
feat: Add LangChainAgentRunner with agentic tool-calling loop
feat: Add OpenAIRunnerFactory.create_agent(config, tools) -> OpenAIAgentRunner
feat: Add LangChainRunnerFactory.create_agent(config, tools) -> LangChainAgentRunner
feat: Add ManagedAgent wrapper holding AgentRunner and LDAIConfigTracker
feat: Add LDAIClient.create_agent() returning ManagedAgent

Requirements

• I have added test coverage for new or changed functionality
• I have followed the repository's pull request submission guidelines
• I have validated my changes against all supported platform versions

Related issues

Provide links to any issues in this repository or elsewhere relating to this pull request.

Describe the solution you've provided

Provide a clear and concise description of what you expect to happen.

Describe alternatives you've considered

Provide a clear and concise description of any alternative solutions or features you've considered.

Additional context

Add any other context about the pull request here.

---

Note

Medium Risk
Introduces a new managed agent execution path (LDAIClient.create_agent) and new provider-side agent runners, which changes how tool definitions, instructions, and token usage are interpreted/tracked. Risk is mainly around compatibility with provider SDKs (langchain / openai-agents) and correct tool/usage wiring rather than data/security concerns.

Overview
Adds first-class managed agent support to the server AI SDK via ManagedAgent and LDAIClient.create_agent(), enabling tracked agent-style invocations (instructions + tool calling) distinct from model and agent-graph flows.

Implements provider-side create_agent support for OpenAI and LangChain by introducing OpenAIAgentRunner (backed by openai-agents) and LangChainAgentRunner (backed by langchain.agents.create_agent), including tool binding/validation and aggregating token usage from agent runs.

Refactors token usage extraction to handle more response shapes (OpenAI chat completions vs agents RunResult, LangChain usage_metadata vs response_metadata) and centralizes native OpenAI tool mapping in openai_helper; updates tests accordingly and exports the new runners from provider packages.

^{Written by Cursor Bugbot for commit 268b669. This will update automatically on new commits. Configure here.}

[keelerm84]

Looks like bugbot has some good feedback on this one.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

[cursor]

Falsy or-chain conflates zero tokens with missing attribute

Low Severity

The or-chain in get_ai_usage_from_response uses Python truthiness to fall through attribute lookups: getattr(usage, 'input_tokens', None) or getattr(usage, 'prompt_tokens', None) or 0. Because 0 is falsy, a legitimate input_tokens = 0 is treated the same as a missing attribute, causing the chain to fall through and check prompt_tokens. If a future SDK version exposes both field names on the same usage object, a zero input_tokens could incorrectly resolve to a non-zero prompt_tokens. Using is None checks instead of or would correctly distinguish "zero" from "absent."