Modularize Preprocessing Pipeline

[nachovizzo]

As exposed in #290 where we made the registration module a class, I think we should go ahead and do the same with the preprocessing module, allowing us for more customization/parametrization of this bit. I'm copying the last part of that PR in here:

From a system point of view, it's nice that we now have this design:

// KISS-ICP pipeline modules
Registration registration_;
VoxelHashMap local_map_;
AdaptiveThreshold adaptive_threshold_;

Makes me wonder if we shouldn't merge Deskew into Preprocessing and convert it to another component as well 👀 , similar to how we do it in Python 🤔 , but to be discussed later

Items to address

• Deskew is a preprocessing step, should be there
• I'd love to move the preprocessing step for Kitti out of our CORE library and just drop it in the Python bindings to avoid people reporting the wrong number in papers
• Decide if it's worth doing the deskewing with tbb, probably we should benchmark it (I think the answer is yes). If we keep it, we should decide on a clean way to control the number of threads "globally" Add support for limiting num_threads in tbb task #252 (comment)

CC @tizianoGuadagnino , @benemer