Fix TODO about covariance in comment

[matt-graham]

The answer to the question in

ParticleDA.jl/src/OptimalFilter.jl

Lines 114 to 122 in c278bbc

	# Covariance between observations R_22, from equationd 3-4 in in Dietrich & Newsam 96
	# TODO: Ask Alex why we add sigma^2 on the diagonal
	function covariance_stations!(cov::AbstractMatrix{T}, grid::Grid, stations::NamedTuple, noise_params::NamedTuple, std::T) where T
	cov .= covariance.(abs.(stations.ist .- stations.ist') .* grid.dx,
	abs.(stations.jst' .- stations.jst) .* grid.dy,
	(noise_params,)) .+ I(stations.nst) .* std.^2
	end

is I believe explained in the first complete paragraph in the second page (page number 1645) of Dietrich and Newsam 1995

Second, in practice, measurements z₂* of Z on Ω are likely to be noisy. This is usually modelled by assuming that [z₂*]ₖ = Z(xₖ, yₖ) + εₖ, where the errors εₖ, are zero mean Gaussian random variables which may be correlated among themselves with correlation matrix Σ, but are independent of the vectors z₁ and z₂. This extension can be easily accommodated in the results presented above by simply replacing the matrix R₂₂ in (3) and subsequent equations by R̅₂₂ = R₂₂ + Σ.

In our case we are assuming Σ = σ²I (that is the observation errors / noise variables are independent zero-mean Gaussian random variables with standard deviation σ) so R̅₂₂ = R₂₂ + σ²I.

If that answers the question I can create a PR to update the comment to summarize the above.