Fix contract tests

[leburgel]

Fixes tests for contract(::AbstractTTN, ::AbstractTTN) by fixing a bug in inner(::AbstractTTN, ::AbstractTTN) and changing the initial states used in the tests. Aside from the bug making it it seem like the algorithm failed when it actually didn't, the updated behavior for the initial state always starts from a product state by default. So the nsite=2 case needs more sweeps to be able to grow the bond dimension enough, and the nsite=1 case couldn't work without explicitly supplying an initial state with a nontrivial bond dimension.

[mtfishman]

Great, thanks @leburgel! Glad it was a simple fix and not even an issue with contract.

That is an issue with the current default for init, fundamentally it is a difficult issue for any of these algorithms (DMRG, TDVP, etc.) to choose general initial guesses and that's way we make people input initial guesses in DMRG. So we'll have to think carefully about that going forward, and maybe just force people to input an initial guess but give tools like forming a random TTN of a certain bond dimension to help with it and also provide tools for extracting the external indices in a form that works with this function, based on the input TTNs being contracted.