Torchvtk

This release fixes more bugs, namely:
- Working with `np.ndarray` inside `TorchDataset`s
- `make_nd` now throws an exception if it's called to reduce Tensor cardinality
- Added `torchvtk.utils.clone()` as a way to clone objects of unknown type (torch.Tensor, np.ndarray, list, tuple, dict, otherwise noop)

Also:
- A sample volume data file is included in the package
- More image / TF plotting tools

**Full Changelog**: https://github.com/torchvtk/torchvtk/compare/0.4.5...v0.4.6

Clone tensor tiles

Adds option to tile only along specific dimensions. Calling `TorchDataset.tile(tile_sz=(None, 128, 128))` will tile the volumes into `128x128` chunks along the X and Y dim, whereas the Z dim always uses the full length

Further a TorchDataset now has the option to return all tiles for a given volume in the non-preloaded case. See below:

ds = TorchDataset('/path/to/ds').tile(tile_sz=(None, 128, 128), return_all_tiles=True)
a = ds[0] This is now a list containing all tiles for volume 0

Added `torchvtk.transforms.RandLambda` to create lambdas with random variables