clarification and fixes for flux2 and other i2i model validation setup

[bghira]

This pull request enhances documentation and adds new model capability methods to support and clarify advanced validation workflows, especially for models that use conditioning inputs (such as reference images or control signals). The main improvements are the addition of two new validation options (--validation_using_datasets and --eval_dataset_id) across all supported languages, and new methods in the model helper to distinguish between required and optional conditioning support.

Documentation improvements:

• Added explanations for --validation_using_datasets and --eval_dataset_id options in all documentation files (OPTIONS.md, .es.md, .pt-BR.md, .zh.md, .hi.md, .ja.md). These options allow validation using images from specific training datasets and enable advanced workflows for models supporting image-to-image or conditioning-based validation. [1] [2] [3] [4] [5] [6]
• Expanded documentation to clarify the three main paradigms for conditioning support (required, optional, and validation modes), including tables of conditioning types and dataset configuration requirements. [1] [2] [3] [4] [5] [6]

Model capability methods:

• Added requires_conditioning_dataset() and supports_conditioning_dataset() methods to simpletuner/helpers/models/common.py. These allow models to declare if conditioning datasets are mandatory or optional, enabling the WebUI and training pipeline to adapt UI and validation logic accordingly.
• Provided detailed comments and usage examples for these methods, improving code clarity and maintainability for future model integrations.