see badges/shields#8671

"TAT.hpp" is used to describe change of TAT in c++ side and "TAT.py" ise used to
describe change of TAT in python side.

1. Remove FastName binding for TAT.py
Name stored in python is always string, it is needed to calculate hash
every time when calling function in c++. So it does not make program
faster to using FastName in python side.

2. Combine navigator.cpp and random.cpp into PyTAT.hpp.
It is not needed to compile them individual to speed up the
compilation.

3. Add binding for some new function in Edge type, and remove binding for
deprecated functions.

4. Update function argument names to keep the same between c++ and python side.

5. Remove old navigator function, use alias to get tensor easily, see below. The
navigator is only used to get the compile and license information now.

6. Add module alias, user can use `TAT.Normal.float64.Tensor` to get normal
float64 tensor instead of old `TAT("No", np.float64)` or original
`TAT.No.D.Tensor`.

7. Getting single block of tensor use the code in mdspan now, instead of
previous duplicated code in TAT.py. It also return numpy array always, instead
buffer protocol.

8. Fix bug when getting arrow of no symmetry edge.

Using safe rename in common_toolkit solves this.

Currently only works for openmpi.