fix: properly closing tcp connections in sdk to fix timeout errors

[Bryce-Soghigian]

Description
We will occasionally get an error like:

{"name":"sm-general-purpose-vdnt7"},"namespa
ce":"","name":"sm-general-purpose-vdnt7","reconcileID":"11d07c7c-6c56-4d0a-b49f-50393126d7a9","error":"launching nodeclaim, creating instanc
e, Get \"https://management.azure.com/subscriptions/(redacted)/providers/Microsoft.Network/locations/eastus/operat
ions/(redacted)?api-version=2022-01-01\": dial tcp 20.62.130.212:443: i/o timeout"}

This is caused by an issue in the go std library Ref
When a TCP connection is dropped, there is a bug in the http2 implementation in go that misses to mark the connection as failed. as a result, the connection stays in the pool, and the calls using that connection hit a context deadline.
The connection can stay in the pool fur up to 17 min (with default linux kernel settings).

This PR updates our http client to share its configuration with the AKS Recommended Default middleware provided in the azure-sdk-for-go-extensions project that fixes the pool getting filled up.

How was this change tested?

• unit tests in the azure-sdk-for-go-extensions
• proven to fix issue in prod via various aks services that have adopted these client options

Does this change impact docs?

• Yes, PR includes docs updates
• Yes, issue opened: #
• No

Release Note

Fixing TCP Timeout Errors from stale connections in the pool 

[Bryce-Soghigian]

/test

[Bryce-Soghigian]

See PR Adding this update to the middleware: Azure/azure-sdk-for-go-extensions#29

[Bryce-Soghigian]

Coveralls is failing because i removed code in a file with 100% coverage so it lowered overall percentage

[Bryce-Soghigian]

/test