Implement transaction search and filtering API

[miss-yusrah]

Closes #98

Transaction Search & Filtering API

Overview
Implements a powerful search and filtering endpoint that allows querying transactions by multiple criteria simultaneously, supporting operations and compliance workflows.

Changes

New Endpoint

• GET /transactions/search - Search transactions with multiple filter support

Query Parameters
All parameters are optional and combine with AND logic:

• status - Filter by transaction status (e.g., completed, pending, failed)
• asset_code - Filter by asset code (e.g., USDC, USD, XLM)
• min_amount - Minimum transaction amount (inclusive)
• max_amount - Maximum transaction amount (inclusive)
• from - Start date for date range filter (ISO 8601 format)
• to - End date for date range filter (ISO 8601 format)
• stellar_account - Filter by Stellar account address
• limit - Results per page (default: 50, max: 100)
• cursor - Pagination cursor from previous response

Response Format

{
  "total": 142,
  "results": [...],
  "next_cursor": "eyJjcmVhdGVkX2F0Ij..."
}

Implementation Details

Files Added

• src/handlers/search.rs - Search endpoint handler with parameter validation
• migrations/20260222000000_transaction_search_indexes.sql - Database indexes for search optimization

Files Modified

• src/db/queries.rs - Added search_transactions() with dynamic query builder
• src/handlers/mod.rs - Registered search module
• src/main.rs - Added search route with pool_manager state
• src/utils/mod.rs - Exported cursor module
• src/db/mod.rs - Exported audit module
• src/lib.rs - Exported utils module

Database Indexes
Added 6 indexes for optimized search performance:

• idx_transactions_asset_code - Asset code filtering
• idx_transactions_created_status - Date range with status
• idx_transactions_amount - Amount range queries
• idx_transactions_created_id - Cursor-based pagination
• idx_transactions_search - Combined status, asset_code, and date queries

Security

• All query parameters use parameterized queries (no SQL injection risk)
• Input validation for dates, amounts, and cursors
• No user input directly interpolated into SQL

Performance

• Uses read replica for search queries (reduces primary DB load)
• Cursor-based pagination for efficient large result set traversal
• Optimized indexes for common search patterns