compute cohen's kappa from confusion matrix

[ericphanson]

currently, we compute many metrics directly from the confusion matrix, but not Cohen's kappa:

Lighthouse.jl/src/metrics.jl

Lines 105 to 127 in 2feb223

	"""
	cohens_kappa(class_count, hard_label_pairs)
	Return `(κ, p₀)` where `κ` is Cohen's kappa and `p₀` percent agreement given
	`class_count` and `hard_label_pairs` (these arguments take the same form as
	their equivalents in [`confusion_matrix`](@ref)).
	"""
	function cohens_kappa(class_count, hard_label_pairs)
	all(issubset(pair, 1:class_count) for pair in hard_label_pairs) ||
	throw(ArgumentError("Unexpected class in `hard_label_pairs`."))
	p₀ = accuracy(confusion_matrix(class_count, hard_label_pairs))
	pₑ = _probability_of_chance_agreement(class_count, hard_label_pairs)
	return _cohens_kappa(p₀, pₑ), p₀
	end
	_cohens_kappa(p₀, pₑ) = (p₀ - pₑ) / (1 - ifelse(pₑ == 1, zero(pₑ), pₑ))
	function _probability_of_chance_agreement(class_count, hard_label_pairs)
	labels_1 = (pair[1] for pair in hard_label_pairs)
	labels_2 = (pair[2] for pair in hard_label_pairs)
	x = sum(k -> count(==(k), labels_1) * count(==(k), labels_2), 1:class_count)
	return x / length(hard_label_pairs)^2
	end

It could clean up some of the data flow to do so. Here is an implementation.

function cohens_kappa_from_confusion_matrix(conf)
    p₀ = accuracy(conf)
    pₑ = probability_of_chance_agreement_from_confusion_matrix(conf)
    return (p₀ - pₑ) / (1 - ifelse(pₑ == 1, zero(pₑ), pₑ))
end

function probability_of_chance_agreement_from_confusion_matrix(conf)
    counts_1 = dropdims(sum(conf; dims=1); dims=1)
    counts_2 = dropdims(sum(conf; dims=2); dims=2)
    n = sum(counts_1)
    @check n == sum(counts_2)
    return dot(counts_1, counts_2) / n^2
end