[Shared] Create Q-Learning config & types (Pydantic v2)

[saviornt]

Description

Implement the foundational configuration and data models for the Q-Learning component using Pydantic v2. This includes hyperparameter models (general + tabular-specific), transition tuples, episode statistics, and checkpoint metadata stubs. All models must be fully typed, documented with Google-style docstrings, and support validation + serialization.

Why: Establishes strong typing and config safety across shared, appliance, and assistant code to prevent drift.

Type

• Task

Focus Area (pick one)

• Shared Utils & Models

Priority

• Critical

Acceptance Criteria

• QLearningHyperparams BaseModel with all core hyperparameters (alpha, gamma, epsilon schedule, etc.)
• TabularQConfig extending it with n_states, n_actions
• Transition and EpisodeStats models with proper typing (Any → constrained where possible)
• All models pass strict MyPy (no Any leaks, full coverage)
• Unit tests cover creation, validation errors, JSON round-trip, and field constraints
• Google-style docstrings on every class and important field
• Files created: shared/src/learning/q_learning/config.py and types.py

Blocker / Dependencies

None

Notes / Links

• Related files: shared/src/learning/q_learning/
• Pydantic v2 best practices (use Field for defaults/constraints)