fix: parse new JSONL bulk format, add missig fields & variants, rate limiting

[voidcontext]

The bulk data format has changed, see: https://scryfall.com/blog/two-new-ways-to-sync-scryfall-data-236

[voidcontext]

@mendess There are a lot of test failures because of Scryfalls's rate limiting. I could just sprinkle a bit of "thread sleep" over the snippets in the doc comments, but that would be extremely ugly. Do you have any better suggestions?

[voidcontext]

@mendess There are a lot of test failures because of Scryfalls's rate limiting. I could just sprinkle a bit of "thread sleep" over the snippets in the doc comments, but that would be extremely ugly. Do you have any better suggestions?

I think I pushed a reasonable solution. Respecting rate limits by default is a good idea probably, and through feature flags it's possible to disable it if needed. This doesn't solve all the rate limiting related test issues, so I'll need to keep digging.

What do you think about this solution? Would it be acceptable?

Edit: obviously this would add an extra 100-500 ms delay to every requests even if it's not required. I'll try to come up with an improved solution.

[voidcontext]

The build is green, all the known issues have been fixed (there's one flaky test though, that failed locally from time to time: search::tests::random_works_with_search_options)

I've made some changes to the test script that helped me during development (e.g. to build all the different feature permutations with scripts/test-features.sh build).

From a developer experience perspective (for contributors of this library), I am not really happy with the end result, because a simple cargo test will run into rate limit issues, but making the blocking feature default would mean that the library users would need to disable the default features to not be needlessly blocked in normal user scenarios. Maybe this should be documented in CONTRIBUTING.md.