Migrate from SQLite to Azure Table Storage for persistent inventory data

[seligj95]

Summary

Migrated the inventory MCP server from SQLite to Azure Table Storage to support the read-only filesystem in Azure Functions deployment environment.

Changes Made

Storage Migration

• Replaced SQLite database with Azure Table Storage for persistent data
• Implemented lazy initialization with deferred retry on startup failures
• JSON serialization for complex types (sizes dictionary)
• Consistent partition key strategy using INVENTORY partition

Authentication & Security

• Configured Managed Identity authentication with explicit client ID
• Proper RBAC configuration for Table Storage access

Dependencies

• Added azure-data-tables for Table Storage SDK
• Added azure-identity for managed identity authentication
• Removed SQLite3 dependencies

Infrastructure Updates

• Updated infra/main.bicep to enable Table Storage (enableTable: true)
• Updated infra/app/api.bicep to configure STORAGE_ACCOUNT_NAME and AZURE_CLIENT_ID environment variables
• Maintained compatibility with existing Flex Consumption plan deployment

Error Handling

• Robust error handling with graceful degradation
• Non-blocking initialization allowing server startup even if Table Storage is temporarily unavailable
• Comprehensive logging for troubleshooting

Key Files Changed

• server.py - Complete migration to Azure Table Storage with managed identity auth
• requirements.txt - Updated dependencies
• infra/main.bicep - Enabled Table Storage feature flag
• infra/app/api.bicep - Added environment variable configuration

Testing Completed

• ✅ All 5 MCP tools (get_inventory, add_item, get_item_by_id, update_item_quantity, search_items) working via HTTP
• ✅ Data persistence verified across function app restarts
• ✅ Managed identity authentication successful in Azure
• ✅ Local development and Azure deployment both tested

Breaking Changes

None - All API contracts remain unchanged. Existing MCP clients continue to work without modifications.