Tensor Support - Kernels

[CPestka]

Tensor Support - Kernels

This Pr adds support for tensors, for all / most(?) CPU kernels that are non 2D- or frame- specific or that wire up some lib like eigen or cblas (let me know if I missed one) and adds some higher dim / tensor specific ones. I omitted the Cuda kernels for now as well, but if somebody is interested I can add those as well.

One kernel that differs a bit from the rest is the aggregate kernel, which is written to work with async i/o, without the need for the compiler to understand what is going on.

A couple of general points in respect to the kernels, which I have not touched:

1. Besides generaly everything being 2D specific currently (fair enough), a lot of things are named MatX or MatrixX, even if they also operate on frames and now also tensors and imo should be renamed to smth like ObjX like is already the case for some of the kernels.
2. It is unclear to me what the "allowed state" of the res ptr passed into most of the kernels is supposed to be. Some kernels check for a nullptr and create a new fitting object if they encounter one, but do not check whether the object is actually "suitable" to be overwritten if it is not a nullptr (e.g. large enough). Some kernels don't check and always create a new obj (which I did here for the tensor kernels as well).

[CPestka]

Just noticed, that I used concepts in the quantize kernel -> added c++20 in cmake. Note sure why it compiled for me without it, gcc-12 should have complained, as gcc-9 did on the CI machine.

[CPestka]

Just noticed in gcc concepts are only supported for >= 10