Add lesion-wise metrics

[valosekj]

This PR adds the lesion-wise metrics (lesion_ppv, lesion_sensitivity, and lesion_f1_score) proposed in ivadomed/MetricsReloaded#3 to our our compute_metrics/compute_metrics_reloaded.py script.

To test the branch, please follow our installation instructions here and then you can run our compute_metrics/compute_metrics_reloaded.py script on your images, for example:

python compute_metrics_reloaded.py
-reference sub-001_T2w_gt.nii.gz 
-prediction sub-001_T2w_prediction.nii.gz 
-metrics lesion_ppv lesion_sensitivity lesion_f1_score

Please also review unit tests in tests/test_compute_metrics_reloaded.py for different scenarios (empty gt and non-empty pred, non-empty gt and non-empty pred with partial overlap, etc.).

Tagging @naga-karthik, @plbenveniste, and @Nilser3, who work with lesions and thus this PR is super relevant to them.

Resolves: #49

[valosekj]

@naga-karthik I fixed lesion PPV to 0.0 in case of non-empty reference and empty prediction in ad23622. The unit tests are now passing!

[valosekj]

The compute_metrics_reloaded.py script has been moved to https://github.com/ivadomed/MetricsReloaded in ivadomed/MetricsReloaded#5 --> not merging