[Robustness] Create Farsi Tokenizer Robustness Dataset

[Malikeh97]

• Create Canonical datasets in Farsi (with similar complexity level and overall format/category to the Turkish version)

• Generate Farsi-specific perturbations for the canonical examples

• Add them the dataset sheet

• Create a HF dataset for the Farsi tokenizer robustness dataset and push to R3 HF space (the associated collection)

[Malikeh97]

Initial version: https://huggingface.co/datasets/Malikeh1375/farsi-tokenizer-robustness-mmlu