Investigate AKS recorded tests timeouts

[vcolin7]

The recorded tests for AKS are currently timing out without much of an error on stderr or stdout. @scbedd suspects there are insufficient resources on the machine the pipeline runs the tests on for all the Test Proxy instances we spawn at this time, so that's something to consider.

We'll also need to re-enable the recorded tests by adding the assets.json file removed in this PR.

[scbedd]

CC @vcolin7

Context

First, some context on the ordering of operations at hand:

• dotnet test collects targeted test assemblies
• for each test class
  • spin a test-proxy instance, await startup and get URL + auto-assigned port (THIS is where timeout is occurring during resource pressure).
    • The proxy is started in single recording mode so that requests don't need x-recording-id injected.
  • for each test (runs in serial)
    • start azmcp instance, assign PROXY_URL to the proxy url from the spun proxy above
    • start playback or recording with the test-proxy
    • run the test, azmcp instance will url rewrite, but that is all
    • stop playback

In conclusion:

• This PR will mitigate the aks issue, but this will recur when we have additional recorded test classes added to the assembly.
• The primary issue is that the test-proxy incurs pressure on the CPU in that initial startup phase, and we timeout grabbing the URL.
• How can we reduce resource pressure?

Proposed solution 1

• Move test-proxy to per-assembly startup
• Change the order of RecordedCommandTestBase to call proxy test start before mcp tool start
• Inject PROXY_RECORDING_ID into the azmcp tool start.
• Modify HttpClientConfigurator to inject the recordingId into the request if PROXY_RECORDING_ID is set.
• I'm concerned we will be blocked due to our handler code here