Added an option to select models based on a reverse ranking

[amjjbonvin]

What does this PR do and why?

Until now the assumptions as that the best models would have the lowest scores, with a default ascending sorting when ranking them. Some scoring function might have a different behaviour (e.g. higher is better).
This PR implements the option to rank/select models in reverse order.
An option was added to both seletop and seletopclust to rank models/clusters in reverse (descending) order meaning in from high to low scores. The default behavior remain unchanged (ascending order, from low to high scores).

Single structure selection

• Added the sort_ascending parameter: true (default) selects best-scoring clusters first; false reverses the order (highest average score first).

Score-based cluster selection

• Added the sort_ascending parameter: true (default) selects best-scoring clusters first; false reverses the order (highest average score first). The reverse option applies only to the score-based ranking, not to size-based ordering.
• rank_clust_order now ranks clusters by their average score (computed over each cluster's best-scoring models via the shared libclust.rank_clusters helper), instead of just sorting the original cluster-rank keys coming from the clustering module.
• Added cluster_score_threshold (default 4): number of best-scoring models per cluster used to compute the cluster's average score. Replaces a previously hard-coded value.
• Exported cluster PDB files are now named following the average-score ranking: cluster_1_model_1.pdb corresponds to the best-scoring cluster (in ascending or descending order). The export loop renumbers clt_rank by its position in the computed ranking.

How was this tested?

Tested by running a test workflow from the docking-protein-protein example, adding sort_ascending=false to the seletop, seletopclust and caprieval stages and manually control that the ranking is indeed reverted.

Integration test for seletopclust adapted for the new option and new test added for seletop.
All tests passing.

AI assistance

AI was used to sort out some error in the seletopclust implementation and in updating the tests.

Checklist

• Tests cover the new and/or changed code
• Documentation updated if needed (also in the haddock3 user-manual
• CHANGELOG.md updated for user-facing changes

Related issues

#1620

[AnnaKravchenko]

If you could add a separate log for when sortby = ‘size’ (I added suggestinos for that) - it would be extremely perfect !