Add Foundry IQ multi-source retrieval sample

[mattgotteiner]

Summary

• add an end-to-end Python sample with one knowledge base referencing both SearchIndexKnowledgeSource and McpServerKnowledgeSource
• use bundled fictional public data plus the public Microsoft Learn MCP server
• verify from linked activity and references records that both sources participated
• document source inclusion, reasoning effort, reranker threshold, runtime, authentication, troubleshooting, and cleanup guidance
• refuse resource-name collisions and restrict cleanup to ETag-bound resources owned by the sample

Validation

• 12 local tests pass
• clean end-to-end deployment completed setup, cross-source answer synthesis, trace verification, data-plane cleanup, and infrastructure deletion
• observed 2 Search activities / 3 Search references and 2 MCP activities / 14 MCP references
• independent cleanup verification found zero remaining indexes, knowledge sources, or knowledge bases
• no secrets or environment-specific identifiers are committed