1116: Extend multivariate scoring vignette with energy vs variogram comparison

[kathsherratt]

Description

The aim of this PR is to lightly expand the multivariate scoring vignette and add a comparison between the energy score and the variogram score on some example data. For me, the problem of autocorrelation by forecast horizon is the most obvious and intuitive/compelling use case for multivariate scores, so wanted to include something on that (as well as the existing vignette with spatial dependence).

This PR relates to #1116 (now closed) and various others related to enenrgy/variogram scores.

Changes to vignettes/scoring-multivariate-forecasts.Rmd:

• Added a note that the example forecasts are for demonstration only and do not represent joint draws from a multivariate distribution.
• Added an example of pooling across forecast horizons (joint_across = c("horizon", "target_end_date")) so each score covers a whole trajectory.
• Added a section explaining why the energy and variogram scores cannot be applied to quantile forecasts, and why lining up quantile levels as pseudo-samples gives a misleading score.
• Added a section comparing the energy and variogram scores, with a worked two-model example (EuroCOVIDhub-ensemble vs epiforecasts-EpiNow2) including a plot of the forecast trajectories, showing how the two scores can disagree and what each rewards.

Also adds a NEWS item.

Checklist

• My PR is based on a package issue and I have explicitly linked it.
• I have included the target issue or issues in the PR title as follows: issue-number: PR title
• I have tested my changes locally (vignette renders cleanly with the dev package loaded).
• I have added or updated unit tests where necessary (N/A — vignette and NEWS only, no code changes).
• I have updated the documentation if required (the vignette is the documentation).
• I have built the package locally and run rebuilt docs using roxygen2 (N/A — no changes to R code or roxygen comments).
• My code follows the established coding standards and I have run lintr::lint_package() to check for style issues introduced by my changes (no lints in the changed vignette).
• I have added a news item linked to this PR.
• I have reviewed CI checks for this PR and addressed them as far as I am able.

Dev: I designed this, implemented, then used iterative LLM support

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.15%. Comparing base (06ed56d) to head (8936cfb).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1193   +/-   ##
=======================================
  Coverage   98.15%   98.15%           
=======================================
  Files          41       41           
  Lines        2225     2225           
=======================================
  Hits         2184     2184           
  Misses         41       41           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[kathsherratt]

I wanted to keep this PR very light, but noting a couple of ideas for further development -

• The vignette pretty much gets straight into functionality without very much on the use case; we could structure the vignette with more intro and rationale for multivariate scoring in general
• Relatedly this could draw in from updated material on multivariate scores developed for teaching by @seabbs @nickreich

[nikosbosse]

Nice, thank you!

Should we warn more explicitly about multivariate quantile scores in the code?