[FEATURE] Define benchmark instances for solving cryptarithms

[arnaud-m]

Is your feature request related to a problem? Please describe.
We need to compare the different models for solving the cryptarithms.

Describe the solution you'd like
We need to define benchmark instances for the following decision problem.
Is the input equation a cryptarithm with a unique solution ?

• The benchmarks instances are divided into thematic datasets.
• The instances are the most difficult. Easy instances are discarded.
• Each dataset is a single file with an instance per line.
• Datasets are generated thanks to the generation model.

Describe alternatives you've considered
Generate the datasets instances on the fly, but it is slower and less reproducible.
Indeed, the candidate set can change depending on the generation model.