📈 Add performance benchmarking suite

Description

Create a benchmarking suite to measure and track performance metrics over time.

Current Status - Partially Implemented

The benchmarking infrastructure exists but needs enhancement for comprehensive performance tracking.

Already Implemented

• ✅ Criterion crate integration in Cargo.toml
• ✅ benches/ directory structure
• ✅ Basic benchmark in benches/inference_benchmark.rs
• ✅ CI job running benchmarks (benchmarks job in ci.yml)

Remaining Work

• Enhance existing benchmarks for ResNet18 built-in model
• Add memory usage benchmarking
• Add throughput benchmarks for different batch sizes
• Add concurrent request benchmarking
• Store benchmark results for comparison
• Add performance regression detection
• Add HTTP endpoint benchmarking (end-to-end)

Updated Implementation Focus

Since the app now uses built-in ResNet18 model:

• Benchmark with 150,528 input elements (3×224×224)
• Test with realistic ImageNet-style input data
• Measure inference time for built-in model vs dummy fallback
• Benchmark batch processing (1, 2, 4, 8 images)

Implementation Guidance

• Enhance existing benches/inference_benchmark.rs
• Use criterion for Rust benchmarking (already configured)
• Add benchmarks for the core inference functions with ResNet18
• Consider using hyperfine for end-to-end HTTP benchmarks
• Benchmark both successful inference and error cases

Estimated Difficulty

Medium - 1 day (reduced due to existing infrastructure)