Rust High-Performance Data Processing Pipeline Design Prompt

You are a senior Rust systems engineer specializing in high-performance data pipelines. Help me design a data processing pipeline with the following requirements: Input: [describe your data source, e.g., PDF documents, log files, API streams] Output: [describe desired output format] Throughput target: [e.g., 10K documents/min] Latency requirement: [e.g., < 100ms per item] Please provide: 1. **Architecture Overview**: Pipeline stages with async channels (tokio mpsc/crossbeam) 2. **Stage Design**: For each stage, specify: - Input/output types (with serde serialization) - Concurrency model (rayon for CPU-bound, tokio for IO-bound) - Error handling strategy (retry, dead-letter queue, circuit breaker) 3. **Memory Management**: Zero-copy parsing where possible, arena allocators for batch processing 4. **Backpressure Handling**: Bounded channels, rate limiting, adaptive batch sizing 5. **Observability**: Metrics (throughput, latency percentiles), tracing spans, health checks 6. **Benchmark Harness**: Criterion.rs benchmark setup for critical hot paths Provide the core Rust code structure with type definitions, trait bounds, and key implementation snippets. Use idiomatic Rust with proper lifetime annotations.