This PR does structural searchs for reasonable TTN. We can try the crossinterpolate function of the version including the structural search.

function crossinterpolate_adaptivetree(
    ::Type{ValueType},
    f,
    localdims::Union{Vector{Int},NTuple{N,Int}},
    g::NamedGraph,
    nsearch:: Int = 10,
    initialpivots::Vector{MultiIndex} = [ones(Int, length(localdims))];
    kwargs...,
) where {ValueType,N}

The variable of nsearch is the number of structural searches that will be conducted.
This only changes the TTN structure by swapping the local two tensors so far. However you can find another proposer in newstractureproposer.jl