Splitting elements is incompatible with vectorization

[Hespe]

The following code currently fails on the master branch:

import torch
import cheetah

drift = cheetah.Drift(length=torch.tensor([1])).broadcast((2,))
drift.split(torch.tensor(0.1))

The stack trace points to

cheetah/cheetah/accelerator/drift.py

Line 77 in b38a654

while remaining > 0:

with message

RuntimeError: Boolean value of Tensor with more than one value is ambiguous

The problem disappears without the call to .broadcast().