entropy may sometimes return NaN

[HumpyBlumpy]

The current entropy function sometimes fails by returning NaN. This occurs for example after using changebonds to increase the bond dimension of a state. Perhaps it is better to use something like this?

function entropy(state::Union{FiniteMPS,MPSComoving,InfiniteMPS},n::Int=0)
    pv = entanglement_spectrum(state,n).^2 .+ 1e-20
    -sum(log.(pv) .* pv)
end

[Jutho]

I guess the error comes only from entanglement values which are exactly zero. Maybe it's better to define a more robust function to compute x*log(x) that deals with x=0 properly, and use this in the definition.

[maartenvd]

I have pushed a potential fix, but I'm not yet completely happy (I want entropy(finitemps) to just work, but currently it complains about the non-existence of similar(CRView)).

I'm against using entanglement_spectrum, as it turns the tensormap into an array, which is not always defined (anyonic symmetries) while the entanglement entropy still is defined in those cases.