supporting generic types

[GiggleLiu]

Supporting generic types in ITensor. Generic types can be used for the following purpose

• High precision tensor network contraction with Bigfloat or DoubleFloat

• Tropical tensor network contraction for solving combinatoric optimization problems. Now the following example works

julia> using ITensors, TropicalNumbers

julia> it = ITensor(TropicalF64, [Index(2) for i=1:3]...)
ITensor ord=3 (dim=2|id=852) (dim=2|id=30) (dim=2|id=299)
NDTensors.Dense{Tropical{Float64},Array{Tropical{Float64},1}}

julia> it * it
ITensor ord=0
NDTensors.Dense{Tropical{Float64},Array{Tropical{Float64},1}}

[GiggleLiu]

@mtfishman Hi, is anyone going to review this PR?

[mtfishman]

@GiggleLiu, for some reason I didn't get a notification about the PR, thanks for the reminder about it. Looks good, glad it is such a minimal change. I'll merge this.

Do other operations work as well? For example, does a full DMRG run work?

[GiggleLiu]

No worries. I haven't tried a full DMRG. I was using it for contracting random tensor networks. I can try DMRG later.

[mtfishman]

Perhaps we should move to using LinearAlgebra's 5-arg mul! instead of an explicit gemm! call to make this more generic (and possibly more efficient, since hopefully for custom element types it fuses all of the operations together without temporary arrays).

No rush, thanks.