Add Specificity

[FlorianMF]

🚀 Feature

In addition to Precision and Recall it would be nice to have a Specificity metric.

For the implementation I think it would be enough to make a copy of Recall (class und function) and adapt numerator and denominator in _precision_compute.

Alternatives

For binary classification Specificity is the same as Recall with 0 as true label.
For multiclass classification this is not as easy as this though.

[arv-77]

@Borda , can I take up this issue?

[arv-77]

@Borda , for the test cases, since there is no specificity metric in sklearn, the randomized test cases cannot be used. Should I just write some test cases with fixed values?