perf(gatsby-source-contentful): experimental flag to skip id normalization

[pvdz]

This adds an experimental flag, EXPERIMENTAL_CONTENTFUL_SKIP_NORMALIZE_IDS, to skip the normalization step when sourcing from Contentful.

This step would take the sys.id and force it to a string and not start with a number. This step is quite expensive at scale, easily taking 5 minutes or more, because it has to traverse every property on every object of the content retrieved by Contentful.

While this step may have been necessary a long time ago, everybody agrees that today it's not required.

We may concretely remove the step in the future, but since that's a breaking change (people can query the contentful_id field in graphql, or access the contentful_id property directly inside a node) that's going to take a little longer to be published. This flag serves as a workaround until that time.

[gatsby-cloud-staging]

Your pull request can be previewed in Gatsby Cloud: https://build-1abee51d-ddc6-4fea-a1a4-2d9242ccd9bf.staging-previews.gtsb.io

[axe312ger]

Has high potential to fix issues and due to the flag we won't break peoples projects. Lets go :)

[xavivars]

@pvdz: Will there be any way to access the contentful_id after this got merged?

I think I originally misunderstood the change. Am I right that contentful_id now is "exactly" the same Id that the contentful entry has, instead of being a modified version of it?

[pvdz]

Correct me if I'm wrong but that should be the .id now, since it's no longer normalized? Before that was "backed up" in contentful_id.

[axe312ger]

id will be a custom id, build from the content type and the id of the entry. (As entrys of different types can share the same id)

The contentful_id will now contain the real id from contentful, without any alteration.