Add similarity-sorted correlation matrix figure to IBMA reports

[tsalo]

Summary

One issue with images on NeuroVault is that the contrast is not always clear or correctly labeled.
I was thinking that something that could help diagnose inverted contrasts would be a similarity-sorted correlation matrix figure, like the one in the NARPS paper that helped identify which processing pipelines gave inverted results to the others? It might be useful for flagging images with the wrong contrast (see below).

I don't necessarily think this would need clusters.

It could alternatively be a diagnostic of some kind.

[JulioAPeraza]

@tsalo Should we show the entire matrix or only the upper triangle?

[tsalo]

I think either would work. The upper triangle probably takes up less space though, so maybe try that first?