Write tests mimicking slow and intermittent network responses

[heaths]

As follow up from discussions, implement some integrations in azure_core that mimic slow, chunky network responses where some of those chunks may exhibit a network error. This point is to make sure that - whether calling Response<T>::into_body() and streaming the ResponseBody or calling Response<T: Model>::deserialize_body() - we correctly retry requests.

Note: this is not about ranged request headers e.g., with Storage, but that we at least start over from the beginning and try again for both retriable HTTP status codes as well as network "hiccups".

[heaths]

See #1140. This was brought up for track 1 code as well and into_body() definitely escapes outside the pipeline. I believe at the very least methods for which we buffer the response should be retried. Ideally, we can also handle that for streaming responses but would likely have to start over except for service methods that support Range headers.