fix: adobe provider - refresh kit meta when font is not resolved

[drevantonder]

Fixes #197

Updates the Adobe provider:

• Updates the kit fetching to allow for bypassing the cache
• Re-fetches the kit if a font is missing

Update tests:

• Adds a test for the above
• Updates the handles invalid JSON from adobe api Might need to be in a separate PR?
  • Restores console.error by calling vi.restoreAllMocks() (Errors whilst I wrote the tests weren't working)
  • Update the mock fetch to return a promise (or TypeError: fetch(...).then is not a function is actually thrown, not the error that is expected)

[danielroe]

this will end up refetching adobe kits on every request for a font family that the kit doesn't have. this gets called a lot, so this doesn't seem to be the right approach.

some ideas to consider (possibly implementing all of them):

• allow adobe to accept an option to automatically 'refresh'
• a timeout/debounce based approach that doesn't refresh any more frequently than x seconds
• cache font families that are not included in the kit even after a refresh so we don't keep refreshing it for them

[drevantonder]

oh, haha, that would be pretty undesirable to have a refetch every request 😅.

As an alternative to the spammy requests, how does this sound?

• Fetch and cache all the available font families by using api/v1/json/libraries/full (with a little pagination)
• In resolveFont, if a font is requested (and at least one kit id is provided so we don't spam non-adobe users) that is not in the kit but is available in Adobe, print a warning like ${fontFamily} is available from Adobe but not in the provided kits. Restart if you've changed a kit or add the font family to a kit and restart.
• When the above occurs, add a flag item (e.g. refetch-adobe-kits.json) to the cache to reload the kit on the next restart if this item exists.

This would:

• Avoid making spammy requests
• Warn the developer if needed and provide instructions to fix

If that sounds sub-optimal, I'm happy to implement the debouncing. I'll implement the above to have a proof of concept.

[drevantonder]

api/v1/json/libraries/full takes a while with pagination (~20sec). I'll attempt the debouncing method.

[drevantonder]

I've implemented a simple debounce mechanism and a negative cache. I've set it currently to 5 minutes between refreshes.

As to a refresh option. I guess I'm biased to not have it as this figuring out that the kits were being cached, where they were being cached, and how long was a pain, but I'm happy to add a refresh option If you think better? (If so, would we want it to default to true or false?)