Stub Responses do not trigger response handlers

[trevorrowe]

There have been multiple issues reported with response stubbing. While the feature works well in the normal case, it have a few notable issues.

The root cause is how the SDK attempts to simulate response stubs. It excersizes the entires HTTP request stack up to, but not including sending the request. This ensures the request is valid and can be marshaled. At this point is by-passes sending the HTTP request and directly returns a response with stubbed data.

The problem with this approach is that it does not exercise any of the response handlers and callbacks. This has caused the following issues:

• stub response does not populate response targets: Stubbed Responses not written via :response_target amazon-archives/aws-sdk-core-ruby#197
• stub responses do not trigger simplified attribute handling for dynamodb: Can't Stub attribute values with Aws::DynamoDB::Client. #770
• stub responses do no handle nils properly: How can I stub response including nil value? #763
• stub responses tigger pagination: Change default :next_token to nil when using stub_responses #804
• stream handling fix breaks stub data: How to stub content-length of S3 object? amazon-archives/aws-sdk-core-ruby#220

The proposed fix is to generate actual HTTP responses from the stub data. This will allow the stub handler to trigger all of the HTTP events and exercise the full stack of handlers.

[trevorrowe]

All known response stubbing issues have been resolved as of v2.1.0 which is now public.