Partially Implements: #3662 (does not move rate limits to after aggregation).
Changes:
Data in an envelope is always considered to be both Indexed and Non-Indexed, for example: Until a transaction is removed from an envelope it is considered to be both Transaction and TransactionIndexed. Only dynamic sampling, when dropping the transaction from the envelope, emits one outcome with the TransactionIndexed category.
A few tests asserted the old/wrong behaviour of only emitting a single outcome, e.g. when an inbound filter filters a transaction we used to only get a Transaction outcome but not a TransactionIndexed, these tests have been aligned.
When checking rate limits for an envelope, quota on the 'base' category (e.g. Transaction) is now consumed.
Metrics now remember whether they were extracted from a sampled envelope item.
Metrics are immediately rate limited using the same rate limits as the ones used for the envelope.
Metrics from sampled envelopes will not be rate limited again.
Improved cached rate limits with a new type struct, which surfaced a bug that they weren't always correctly expired (fixed).
Future Improvements:
Move metrics rate limiting to after aggregation
Stream line metrics rate limiting from 4 (cached transactions/spans, transactions/spans, cached metric_buckets, metric_bucket) different checks to 2 (cached, non-cached)
Partially Implements: #3662 (does not move rate limits to after aggregation).
Changes:
TransactionandTransactionIndexed. Only dynamic sampling, when dropping the transaction from the envelope, emits one outcome with theTransactionIndexedcategory.Transactionoutcome but not aTransactionIndexed, these tests have been aligned.Transaction) is now consumed.Future Improvements: