[Appliance Core] Implement dummy environment & basic Q-Learning agent

[saviornt]

Description

Build the initial Appliance-side Q-Learning agent and a dummy network environment so full episodes can be run locally on the Pi (before real telemetry integration).

Why: Validates agent composition with shared core and allows early loop testing.

Type

• Feature

Focus Area (pick one)

• Appliance Core (Pi edge)

Priority

• Critical

Acceptance Criteria

• DummyNetworkEnv with discrete state/action space and basic reward function
• ApplianceQAgent composes shared TabularQ + replay buffer + config
• Runs 5000+ steps/episodes end-to-end without crashing
• Logs episode stats (total reward, length, avg reward)
• Uses only shared types/config
• Simple test script shows reward trend improvement over episodes

Blocker / Dependencies

• Phase 1 shared core issues

Notes / Links

• Target files: appliance/src/learning/q_learning/agent.py and dummy_env.py