count as TP only when there's at least 10% overlap

[naga-karthik]

This PR updates the lesion_f1_score metric by modifying the condition for a "true positive" prediction. Previously, it was this here:

Masks are considered true positives if at least one voxel overlaps between the truth and the prediction.

but taking only 1 voxel might be a bit too less so I updated the code to consider at least a 10% overlap to assign a prediction as "True Positive". Note that this is in line with how anima also assigns true positives.

output of animaSegPerfAnalyzer -h

   -z <double>,  --MaxFalsePositiveRatioModerator <double>
     Percentage of the regions overlapping the tested lesion is not too
     much outside of this lesion. (default 0.65)

   -y <double>,  --MaxFalsePositiveRatio <double>
     Maximum of false positive ratio to limit the detection of a lesion in
     GT if a lesion in the image is too big. (default 0.7)

   -x <double>,  --MinOverlapRatio <double>
     Minimum overlap ratio to say if a lesion of the GT is detected.
     (default 0.10)

   -v <double>,  --MinLesionVolume <double>
     Min volume of lesion for "Lesions detection metrics" in mm^3 (default
     3mm^3).

I did not touch the False Positive part yet.

Any thoughts, @valosekj @plbenveniste ?