Remove tests that require User-Agent and x-ms-client-request-id override

[antkmsft]

Please help me to find out the truth: either all SDK should implement conformance to these 2 scenarios, or it should not be allowed - so then the scenarios should be removed, and these SDKs who have logic to make it work probably need to undo the code that they have to enable it.

• Java, .NET, and Python do allow to override User-Agent, C++ and Go do not;
• .NET, Python, and Go do allow to override x-ms-client-request-id, C++ does not.

I am having several conversations with people saying that User-Agent and x-ms-client-request-id overriding should not be allowed:

• Telemetry policy should not be setting User-Agent if it is already set azure-sdk-for-go#19890
• Policies do not overwrite HTTP headers if they are already set azure-sdk-for-cpp#4303 (comment)

[ahsonkhan]

@mikekistler, @lmazuel could you help clarify what the consistent behavior should be?

[mikekistler]

I think this is a question for the SDK architects to decide, but I can give my opinion.

• User-Agent: I think we should allow users to override but also send x-ms-useragent with the original contents of user-agent so we can still get telemetry.
• x-ms-client-request-id: I think we should let users specify this value. It is, after all, for "clients".

[jhendrixMSFT]

The vast majority of these tests were written during track 1 and have not been updated to reflect our current thinking.

This test specifies user-agent in its OpenAPI spec as an input parameter which I believe is something we'd never allow in a production API. So already I think the test is suspect.

[lmazuel]

Customers should be allowed to override both, but not through Swagger, through Azure-Core features.

IOW, I would agree to remove the "User-Agent" parameter from the Swagger, as it should not be necessary to trigger an override. But the test itself checking that people can override the value should stay.