Add `scalar_spread`

[ph-kev]

For the calibration pipeline in ClimaCoupler, this function is used at the end of iteration in analyze_iteration. This would be an useful function to have in general.

function scalar_spread(ekp)
    g_mean_final = EKP.get_g_mean_final(ekp)
    g_final = EKP.get_g_final(ekp)
    sq_dists = [sum((col .- g_mean_final) .^ 2) for col in eachcol(g_final)]
    return Statistics.mean(sq_dists)
end

[ph-kev]

It is not clear to me what this mathematically represent. I thought this was equivalent to computing the standard deviation, so I tried Statistics.std(g_final; dims = 2), but this computes the standard deviation across the same entries of different column.