Unify `TTNR` with `vertical_stack_linear`?

[Yue-Zhengyuan]

I just realized that TTNR is just the (2+1)D generalization of vertical_stack_linear introduced in #189, and want to ask opinions to somehow unify these two.

In my opinion, TTNR is just the preparatory step to obtain a 2D partition function tensor. The actual 2D RG process happens later. So I intend to make TTNR just a different dispatch of vertical_stack_linear.

[Yue-Zhengyuan]

Ultimately, vertical_stack_linear should be replaced by something in MPSKit, and TTNR belongs to PEPSKit. We can make them test dependencies, while the main package stays independent.