[Benchmarking Framework] Create a standardized form testing dataset

[zdeveloper]

Create a standardized form testing dataset based on a readily available dataset, this will be used for building a standard benchmark for ReportVision

Acceptance Criteria

• The dataset should include at least 100 different variations of the same form
• The dataset should include around 10 different forms(ex: quest form,tb form, syph form etc.)
• The dataset should include the ground truth of the data (label/value) present in each form
• add dataset to the git repo, and whatever code was used to generate it
• name this dataset: reportvision-dataset-1

Additional context
Example of datasets came be coming from NIST or huggingface
Please be very communicative and the AC can be adjusted

[zdeveloper]

potential datasets:
medical reports: https://huggingface.co/datasets/AnubhutiBhardwaj/medical-reports-demo
medical prescriptions: https://huggingface.co/datasets/Technoculture/medical-prescriptions
fake tax forms: https://huggingface.co/datasets/singhsays/fake-w2-us-tax-form-dataset

[schreiaj]

Per discussion - the goal of this is to build a dataset to test the OCR not alignment.