[FEATURE] Investigate techniques for enhancing generated videos

[ncudlenco]

Is your feature request related to a problem? Please describe.
Currently, the quality and realism of generated videos in the simulation are limited by the rendering pipeline and available assets. Modern video-to-video (vid2vid) techniques have the potential to substantially improve the visual fidelity, style transfer, and synthesis quality of simulation outputs, enabling more compelling datasets and research demos. However the resulting videos will not preserve pixel-level fidelity and the segmentation / depth / pose masks will lose their value with this approach.

Describe the solution you'd like
Investigate state-of-the-art vid2vid approaches for enhancing simulation-generated videos. The investigation should include:

• Survey of recent vid2vid research, frameworks, and libraries
• Evaluation of integration feasibility with the current mta-sim pipeline
• Comparative analysis of output quality, resource requirements, and scalability
• Recommendations for one or more approaches to prototype and integrate
• Identify potential risks, licensing constraints, and necessary dataset adaptations
• Evaluate the alternative of installing MTA:SA compatible mods (e.g., graphics, weather, effects, or realism mods) directly in the simulation environment instead of post-processing the output videos

Describe alternatives you've considered

• Relying exclusively on the built-in rendering engine (limits realism)
• Manual post-processing or video editing (not scalable or automated)
• Installing MTA:SA compatible mods for enhanced rendering and effects

Acceptance Criteria

• Survey of state-of-the-art vid2vid methods is documented
• Feasibility of integrating selected methods with mta-sim is evaluated
• Comparative analysis of quality, resource usage, scalability, and risks is provided
• Recommendations for prototyping and integration are documented
• Evaluation and documentation of installing MTA:SA compatible mods as an alternative
• Documentation and references are updated
• No existing functionality is broken

Additional context

• The outcome could support dataset enrichment, research, and demo creation.
• Consider the trade-offs between post-processing (vid2vid) and in-game mod-based enhancement strategies.