[FEATURE] Expand/collapse GEST graph nodes for granularity control

[ncudlenco]

Is your feature request related to a problem? Please describe.
Currently, there is no mechanism in mta-sim's GEST space to expand or collapse graph nodes (or collections of nodes) to more granular or more generic representations. This limits the ability to move between high-level narrative descriptions and low-level, game-executable sequences, which is necessary for simulation fidelity, authoring, and automated annotation.

Describe the solution you'd like
Implement a feature that allows users or systems to expand/collapse a GEST graph node or a collection of nodes. This should:

• Enable transitioning between high-level concepts (e.g., "Michael's stormy evening") and low-level, game-primitive actions (e.g., walk to window, look outside, hug Mary, etc.).
• Support both single-node and multi-node expansion/collapse, based on context and user/system needs.
• Ensure the lowest level of granularity maps exactly to available game animations/actions in the simulation.
• Allow for rules/configuration to define how nodes are expanded/collapsed.
• Preserve graph temporal and logical relationships during transformation.

Describe alternatives you've considered

• Manual rewriting of scenes at different granularities (labor-intensive and error-prone)
• Fixed-level GEST graphs (not flexible for authoring or simulation)

Acceptance Criteria

• GEST graph nodes/collections can be expanded to lower-level, granular nodes
• Nodes can be collapsed to higher-level, generic representations
• Expansion/collapse rules are configurable and documented
• Lowest granularity is aligned with available simulation/game actions and animations
• Temporal and logical relationships are preserved during transformation
• Documentation and usage instructions are provided
• Tests verify graph transformation correctness and reversibility

Additional context

• Example: "eating in the kitchen" expands to: go to kitchen chair, sit down, pick up food, eat food, stand up.
• Useful for scenario authoring, dataset annotation, and bridging narrative with simulation.