Failure to handle nothing output from Zygote.jl

[MilesCranmer]

julia> using Zygote, DifferentiationInterface

julia> safe_log(x) = x > zero(x) ? log(x) : convert(typeof(x), NaN)
safe_log (generic function with 1 method)

julia> derivative(safe_log, AutoZygote(), 0.0)
ERROR: MethodError: no method matching iterate(::Nothing)
The function `iterate` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  iterate(::Combinatorics.Combinations)
   @ Combinatorics ~/.julia/packages/Combinatorics/Udg6X/src/combinations.jl:13
  iterate(::Combinatorics.Combinations, ::Any)
   @ Combinatorics ~/.julia/packages/Combinatorics/Udg6X/src/combinations.jl:13
  iterate(::Tables.DictRowTable)
   @ Tables ~/.julia/packages/Tables/8p03y/src/dicts.jl:122
  ...

Stacktrace:
 [1] dot(x::Float64, y::Nothing)
   @ LinearAlgebra ~/.julia/juliaup/julia-1.11.2+0.aarch64.apple.darwin14/share/julia/stdlib/v1.11/LinearAlgebra/src/generic.jl:855
 [2] 
   @ DifferentiationInterface ~/.julia/packages/DifferentiationInterface/gjT8p/src/first_order/pushforward.jl:161
 [3] #6
   @ ~/.julia/packages/DifferentiationInterface/gjT8p/src/first_order/pushforward.jl:192 [inlined]
 [4] ntuple
   @ ./ntuple.jl:48 [inlined]
 [5] value_and_pushforward
   @ ~/.julia/packages/DifferentiationInterface/gjT8p/src/first_order/pushforward.jl:191 [inlined]
 [6] pushforward
   @ ~/.julia/packages/DifferentiationInterface/gjT8p/src/first_order/pushforward.jl:220 [inlined]
 [7] derivative
   @ ~/.julia/packages/DifferentiationInterface/gjT8p/src/first_order/derivative.jl:124 [inlined]
 [8] derivative(::typeof(safe_log), ::AutoZygote, ::Float64)
   @ DifferentiationInterface ~/.julia/packages/DifferentiationInterface/gjT8p/src/fallbacks/no_prep.jl:49
 [9] top-level scope
   @ REPL[13]:1
Some type information was truncated. Use `show(err)` to see complete types.

Zygote.jl returns a nothing given an invalid input, so I think this should be correctly handled. Returning a NaN seems like an option?

For AutoForwardDiff, this gives us:

julia> derivative(safe_log, AutoForwardDiff(), 0.0)
0.0