Incorporate azure-sdk / azure-cli into pyperformance suite

[ericsnowcurrently]

This would let us run profiling and benchmarks against workloads involving the azure SDK. I expect our best option would be using the test suite of either repo.

• decide on initial workload to target
• clone the repo locally
• figure out how to run the tests (our proxy for the workload)
• figure out how to add a benchmark to pyperformance
• add the basic benchmark (it is slow so we won't include it in the default set)
• add an equivalent fast benchmark
• make an upstream PR - Add a benchmark for the Azure CLI. python/pyperformance#90
• verify that the fast benchmark is a good enough proxy for our target workload

Later (getting profiling set up takes precedence):

• add a similar benchmark for azure-sdk
• add a benchmark using Azure functions

[ericsnowcurrently]

• https://github.com/Azure/azure-sdk-for-python
• https://github.com/Azure/azure-cli

[gvanrossum]

Probably related: Anthony Shaw reminded me that new Azure Functions users get 1,000,000 calls for free. They are currently using Python 3.6, 3.7, 3.8 in Docker containers (vanilla CPython).

A reduction in CPU time would go straight to the Microsoft bottom line.

There is a Docker image of the Azure Functions runtime so you could use that to benchmark any optimizations. https://docs.microsoft.com/en-us/azure/azure-functions/functions-create-function-linux-custom-image?tabs=bash%2Cportal&pivots=programming-language-python

[ericsnowcurrently]

This is blocked until Azure/azure-cli#16611 is resolved.

[ericsnowcurrently]

This probably isn't critical for now so I'm moving this back to "to do".

[markshannon]

Is this going to happen, or should we close this?