feat: optimize bundle size, FlatList performance, Soroban storage, and ML inference

[Ebuka321]

Summary

This PR implements optimizations across 4 areas:

1. Mobile Bundle Size Optimization (Closes #593)

• Code splitting via Metro experimentalImportBundleSupport
• React.lazy + Suspense for all screen components with fallback skeletons
• LazyScreen component with error boundary for chunk load failures
• sideEffects: false in package.json for tree shaking
• Bundle analysis CI workflow
• Initial chunk reduced to <8MB target

2. FlatList Performance Optimization (Closes #599)

• OptimizedFlatList reusable component with:
  • getItemLayout for O(1) fixed-height scrolling (84px)
  • Optimized windowing: windowSize=10, maxToRenderPerBatch=5, initialNumToRender=10
  • maintainVisibleContentPosition for scroll position preservation
• React.memo on SubscriptionListItem with shallow comparison of relevant fields
• React.memo on InvoiceListItem with custom comparator
• useMemo for list data transformations (sorting, grouping)
• Migrated SubscriptionList from ScrollView to FlatList-based rendering

3. Soroban Contract Storage Optimization (Closes #596)

• Merkle tree library in contracts/utils/src/merkle.rs with:
  • batch_insert / batch_get for batched storage operations
  • generate_merkle_proof and verify for off-chain/cross-contract proofs
  • Non-existent key handling with proof of non-inclusion
• Batch read function batch_get_storage in main contract
• Backward compatible: single-entry reads still work
• Gas benchmark: batch read of 100 entries targets <200K gas (from 500K)

4. ML Model Inference Optimization (Closes #602)

• PyTorch → ONNX export script with dynamic axes for variable batch sizes
• INT8 post-training quantization with 1000-sample calibration
• ONNX Runtime inference server (FastAPI) with:
  • CUDA/CPU provider fallback
  • Request batching support
  • Health check endpoint
• Kubernetes deployment with resource limits (256Mi request, 512Mi limit)
• Accuracy regression test suite comparing ONNX vs PyTorch (F1 within 1%)
• Quantization pipeline script

Closes #593
Closes #599
Closes #596
Closes #602

[drips-wave]

@Ebuka321 Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits