Charge perturbations leading to negative charge values

[sevmag]

Describe the bug
When using the perturbation functionality of the DataRepresentation on the charge values, you can sometimes get a negative charge, which leads to issues in the detector standardization, which uses log10 for most of the detector classes.

graphnet/src/graphnet/models/data_representation/data_representation.py

Line 21 in 368927d

perturbation_dict: Optional[Dict[str, float]] = None,

graphnet/src/graphnet/models/detector/nubench.py

Lines 40 to 41 in 368927d

	def _charge(self, x: torch.tensor) -> torch.tensor:
	return torch.log10(1 + x)

graphnet/src/graphnet/models/detector/icecube.py

Lines 41 to 42 in 368927d

	def _charge(self, x: torch.tensor) -> torch.tensor:
	return torch.log10(x)

Expected behavior
Lower-bound the charge to 0, as this is the only physically meaningful way