Write tests to check score for exact_duplicates and near_duplicates issues

The tests should ensure that the scoring follows this logic
```
If image is not (near/exact) duplicated:
    score=1
else:
    score = 1 / (number of images in the duplicated set to which the image belongs)
```