Add the non-in-place counterpart of `Batch.to_torch`

[dantp-ai]

Currently Batch.to_torch is a in-place operation. A non-in-place counterpart to it should be implemented that returns a new Batch (cf. similar methods like Batch.to_numpy).

Note that the two methods should be named appropriately according to naming conventions in the existing codebase, i.e. in-place methods have a _ appended to their name.