Motivation

While implementing the AWS_PROXY integration with #11167, I've stumbled onto a weird issue: AWS would raise an INTEGRATION_FAILURE error, which the default status code should be 504 (Gateway Timeout, that's already suspicious), but instead returned 500.

After a lot of investigation with @cloutierMat, we came to the conclusion that the CRUD later (PutGatewayResponse, GetGatewayResponse etc) is not in line with the actual behavior, and that the actual default status code is 500, but the CRUD calls still returns 504.

This PR implements the logic to enable the difference between the CRUD layer and the actual invocation layer by attaching a default status code to the exception, while still using the previous default for other exceptions, allowing us to override the behavior.

This behavior is validated by tests.aws.services.apigateway.test_apigateway_lambda.test_lambda_aws_proxy_integration_non_post_method.

Changes

• use the Python exception status code if provided
• add a fix in the provider to deal with the inconsistency between what the AWS CRUD operations returns and what AWS actually does under the hood