Add offline fallbacks, orchestrator placement strategies, and regression tests

[patriziodazzi]

Motivation

• Make the simulator runnable in offline or restricted environments by providing minimal local fallbacks for common third-party libs and packaging metadata.
• Improve orchestrator behavior and placement flexibility to reduce deployment failures and enable a configurable placement strategy.
• Fix several device/orchestrator bugs that caused incorrect resource checks, image handling and microservice lifecycle management.

Description

• Add local fallback implementations for networkx, simpy, and matplotlib under src/ and include a simple maxflow and compose operator used by the simulator.
• Introduce packaging via pyproject.toml and a .gitignore for development artifacts and caches.
• Extend orchestrators: implement ContinuumOrchestrator.deploy, add DomainOrchestrator placement strategy support (first_fit and best_fit), add _select_target_device to pick best-fit by memory, and tighten image-finding/selection and retrieval logic.
• Fix device-level logic in device/base.py by making storage checks inclusive (>=), making store_image idempotent, guarding duplicate retrievals, changing start_microservice to return bool, and correcting terminate_microservice to use microservice id keys.
• Add demo loader adjustments in test/main.py and a comprehensive test/test_regressions.py suite that covers device behavior, orchestrator deployment, continuum deploy, failure cases, and placement strategy correctness.

Testing

• Ran the unit tests with python -m unittest discover -s test -p 'test_regressions.py' -v and the test suite passed.
• Executed the demo with python test/main.py to verify offline import path handling and basic simulation flow, which produced the expected runtime output.

---

Codex Task