fix(memory): implement data source type detection for multiple data source support

[cagataycali]

Overview

This PR fixes the memory tool data source type detection bug reported in issue #90.

Problem

The memory tool was hardcoded to assume 'CUSTOM' data source type, causing failures when users had multiple data sources where the first one wasn't CUSTOM type (e.g., S3 data sources).

Solution

• Added method that properly identifies data source types
• Prioritizes CUSTOM data sources but falls back to first available if none found
• Updated all operations (store_document, delete_document, get_document) to use proper type detection
• Enhanced error messages for unsupported data source types with clear guidance
• Updated tests to mock the new get_data_source API calls

Key Improvements

1. Robust Data Source Discovery: Searches through ALL data sources, not just the first
2. Type Validation: Explicitly checks each data source's type configuration
3. Clear Error Messages: Users get specific errors for S3 and other unsupported types
4. Backward Compatibility: Existing CUSTOM-only setups continue to work seamlessly

Testing

• ✅ All existing unit tests pass
• ✅ Added test coverage for new data source detection logic
• ✅ Pre-commit hooks (formatting, linting, type checking) all pass

Closes

Fixes #90

References

• Original issue: [BUG] memory tool fail to store kb if user have 2 datasources and first one is not CUSTOM #90
• Follows pattern from agent-builder store-in-kb tool as suggested in issue comments