fix(bake): lightning deploys populate gdoc metadata

[marcelgerber]

Fixes #2906.

❗ I didn't test this in any way yet!

The underlying issue here was that gdoc properties like linkedDocuments and imageMetadata didn't get populated in the lightning deploy code path (bakeGDocPost).

I instead extended bakeGDocPosts to optionally take an array of slugs that it should bake, and bake every published gdoc otherwise.

✅ In order to reenable lightning deploys, we will also have to revert ab2b183.

[ikesau]

Thanks for this @marcelgerber !

I got rid of one last snag which is that posts with a new {.image} would be considered as only having updates to their body, thus triggering a lightning update (even though they would need new images to be baked)

So by correctly populating the before & after objects, and including imageMetadata in the list of properties to diff when determining if the update is a lightning update, we now trigger bakes correctly in all cases.

I tested this by running a local deploy server and trying various permutations:

• Republish article with content changes only → only article is rebaked
• Republish article with image, content changes only → only article is rebaked
• Republish article with new image referenced → full bake
• Republish article by removing an image → full bake
• Start deploy server with two lightning updates in .queue → only both articles rebaked
• Start deploy server with one lightning update and one regular update in .queue → full bake

The bolded case is not ideal behaviour, technically we only need to rebake the article there, but I don't think it's that common a usecase and the simpler we can keep lightning updates, the better IMO 🙂

[marcelgerber]

Nice, thank you!

[ikesau]

Feel free to double check my working 😉