Create a comprehensive example demonstrating Flash's load-balancer endpoints with custom HTTP routes. This example shows how to create low-latency APIs using the @Remote decorator with method and path parameters.

Key additions:
- GPU service for compute-intensive operations
- CPU service for data processing
- Pydantic models for input validation
- Comprehensive documentation with cost estimates and troubleshooting
- LiveLoadBalancer for local development
- LoadBalancerSlsResource configuration for production

Improvements made:
- Added pyproject.toml with project metadata
- Added input validation using Pydantic models
- Added Cost Estimates section to README
- Added Troubleshooting section to README
- Updated root README.md to include new example
- Updated 03_advanced_workers/README.md to document the example

All code passes quality checks: formatting, linting, and type checking.

Improvements:
- Remove redundant validation in GPU worker (Pydantic already validates)
- Use UTC timezone for all timestamps instead of local timezone
- Extract VALID_OPERATIONS to module-level constant in CPU worker
- Clarify port documentation in README

Changes:
- GPU endpoint.py: Remove redundant empty list check, use UTC timestamps
- CPU endpoint.py: Add VALID_OPERATIONS constant, use UTC timestamps, simplify operation validation
- README.md: Clarify port differences between standalone (8000) and unified app (8888)

All changes pass quality checks.

Changes:
- Use ComputeRequest model as parameter in compute_intensive() function for proper type safety and validation
- Update GPU router to pass full request object instead of extracting numbers
- Update test code to create ComputeRequest object properly

This ensures consistent use of Pydantic models for validation throughout the GPU worker endpoint.

All changes pass quality checks.

…unction

Remote functions with load-balancer decorators don't have access to custom
types like Pydantic models in their execution context. Move ComputeRequest
to the router module and convert it to a dict before passing to the remote
function, matching the pattern used in the CPU worker.