Binary focal loss: Use pre-computed probs and increase readability

[klieret]

Changes

• Use pre-computed probs_pos and probs_neg rather than calling sigmoid function again (performance boost by skipping two calls to torch.sigmoid)
• Standardize use of tensor method vs torch function to increase readability (make symmetry in code more apparent)

Type of change

Performance / code readability

• 📚 Documentation Update
• 🧪 Tests Cases
• 🐞 Bug fix (non-breaking change which fixes an issue)
• 🔬 New feature (non-breaking change which adds functionality)
• 🚨 Breaking change (fix or feature that would cause existing functionality to not work as expected)
• 📝 This change requires a documentation update

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Did you update CHANGELOG in case of a major change?