feat: Add TypeScript multi-agent A2A sample using the Claude Agent SDK

### Proposal
Add a use-case sample under `02-use-cases/01-conversational-agents/` demonstrating a multi-agent system built with the **Claude Agent SDK in TypeScript**, using the **A2A protocol** between agents on AgentCore Runtime and **AgentCore Gateway** for tools, authenticated end-to-end with **IAM/SigV4**.

### Gap it fills
The repo currently has no TypeScript A2A sample and no Claude Agent SDK A2A sample in any language:
- The existing [`A2A-multi-agent-incident-response`](https://github.com/awslabs/agentcore-samples/tree/main/02-use-cases/01-conversational-agents/A2A-multi-agent-incident-response) use case is Python (Google ADK / Strands / OpenAI Agents SDK) authenticated via Cognito OAuth.
- The `03-integrations/agentic-frameworks/claude-agent/` samples are Python and single-agent (no A2A).

The proposed sample is the TypeScript counterpart with a different auth architecture (SigV4 rather than OAuth) and adds Gateway tool integration.

### Scenario
A DevOps incident-triage copilot: a lead agent (HTTP protocol) delegates over A2A to a log-analyst worker and a runbook worker; the runbook worker looks up service ownership and runbook steps through an MCP tool exposed by AgentCore Gateway (Lambda target).

### Note on SDK coverage
The AgentCore **Python** SDK ships A2A hosting (`serve_a2a`); the **TypeScript** SDK (`bedrock-agentcore` 0.4.x) does not yet. The sample includes two small, scenario-agnostic packages that fill this gap (a Claude Agent SDK ↔ A2A executor bridge and an in-process SigV4 MCP proxy), structured so they could inform a future TS SDK feature. Happy to adjust scope or placement based on maintainer preference.

I have a working, end-to-end-tested implementation ready to submit as a PR.

Start in 02-use-cases/01-conversational-agents/ and compare the existing A2A-multi-agent-incident-response sample with the Claude Agent SDK samples under 03-integrations/agentic-frameworks/claude-agent/. Review the proposed A2A executor bridge, SigV4 MCP proxy, Runtime agents, and Gateway integration; done means a TypeScript incident-triage sample works end to end with IAM/SigV4 authentication.

Written by the indexing model from the issue text.