Motivation

Following botocore update regarding exception handling of SQS (introduced in boto/botocore#3054, released with 1.31.85), we needed to refresh snapshots for the newly added field.
It put into light some changes regarding error codes, so I've adapted the exceptions raised and their mapping in the serializer.
It also showed that the error message is different for QueueDoesNotExist depending on query or json protocol, put into light with a new snapshot test.

Changes

• refresh all SQS snapshots
• modify the exception classes to match their new names
• add those new exceptions to the map for their old names (different between Error and QueryErrorCode)
• added some snapshot tests to validate behavior
• added a quick fix in the provider to have a different message depending on the service name. Might need to think of a better handling of this, but historically error message were the responsibility of providers, not serializer.
• reverted the change in the CloudFormation models regarding using the boto client exceptions.
• edit: I forgot, but I also fixed the marker detection for only_localstack so that we don't run those tests against AWS. I believe one nice step also would be to not create a snapshot entry in the json file if the test does not have the snapshot fixture, I'm sure we could detect that? @dominikschubert
• update the call to connect_to in the query API to use sqs_query client