Take advantage of symmetry in `hessian`

[gdalle]

As the benchmarks of HyperHessians.jl suggest (@KristofferC), symmetry can probably speed up Hessians quite a bit, even though it is not the only factor. It would be nice to take advantage of it here!

[KristofferC]

You won't get this clean code

ForwardDiff.jl/src/hessian.jl

Lines 31 to 37 in 569af35

	function hessian!(result::AbstractArray, f::F, x::AbstractArray, cfg::HessianConfig{T} = HessianConfig(f, x), ::Val{CHK}=Val{true}()) where {F,T,CHK}
	require_one_based_indexing(result, x)
	CHK && checktag(T, f, x)
	∇f = y -> gradient(f, y, cfg.gradient_config, Val{false}())
	jacobian!(result, ∇f, x, cfg.jacobian_config, Val{false}())
	return result
	end

but it is probably worth doing anyway.

[gdalle]

Worst-case scenario, we hide it behind a config option so that it is non-breaking and optional

[KristofferC]

I don't see why it would be breaking.