What changed

This PR refactors OpenDT from a bit of a monolithic app into a microservices setup. The codebase is now organized into four separate Docker services (dashboard, dc-mock, sim-worker, kafka-init) plus a shared library for common code. Each service has its own Dockerfile and runs independently. All services have documentation with READMEs explaining how they work. The shared library (libs/common/) contains all the Pydantic models, Kafka utilities, and config management code that multiple services need.

I also added a centralized config system using YAML files instead of using in-code constants at several places. There's a main config/default.yaml for global settings and separate configs for different experiments in config/experiments/. This makes it easier to manage different simulation scenarios.

A new feature I added is the experiment management system. You can now run experiments with a simple make experiment name=experiment_1 command, and it automatically handles output directories, result tracking in Parquet files, and generation of plots comparing actual vs simulated power consumption. There's also a debug mode that saves all the OpenDC input/output files if you need to troubleshoot.

The simulation engine got some nice upgrades. It now caches results to avoid re-running identical simulations and uses heartbeat messages for reliable window processing.

Finally, I added a Makefile with commands for everything - make up to start services, make logs-sim-worker to view logs, make experiment to run experiments, etc. Services support hot reload during development, so you don't need to rebuild containers every time you change code.