Improve testing of gallery pipeline

[Pennycook]

Feature/behavior summary

#37 introduced a GitHub Action to run the gallery, which will make it easier for us to inspect the images generated after certain changes. If we could find a way to automatically identify whether any of the generated images have meaningfully changed, that would save us the hassle of having to check the images manually.

Request attributes

• Would this be a refactor of existing code?
• Does this proposal require new package dependencies?
• Would this change break backwards compatibility?

Related issues

No response

Solution description

@laserkelvin suggested that we might be able to use a "perceptual hash" to compare the generated graphs against previously generated graphs, as it may be more robust than something like MD5/SHA-512 if something moves a little bit.

Additional notes

We probably wouldn't want the action to fail just because an image changed (since it may be deliberate). Some sort of warning directing our attention to a files that should be double-checked by a human would be good enough.