{Misc} Fix timer for __main__.py

[jiasli]

Description

Fix the incorrect timing logic. Previous it only takes invoke duration into account, while forgetting about the init duration.

This PR fixes the timing logic to include both init and invoke duration.

Testing Guide

Before:

> az version -h --verbose
...
command ran in 0.881 seconds.

After:

> az version -h --verbose
...
Command ran in 1.679 seconds (init: 0.824, invoke: 0.855)

[yonzhan]

add to S171

[jiasli]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 2 pipeline(s).

[jiasli]

Because azdev doesn't honor azure-cli repo's .flake8 (Azure/azure-cli-dev-tools#209), disabling E402 doesn't work in the CI and causes the failure:

/home/vsts/work/1/s/src/azure-cli/azure/cli/__main__.py:13:1: E402 module level import not at top of file
/home/vsts/work/1/s/src/azure-cli/azure/cli/__main__.py:14:1: E402 module level import not at top of file
/home/vsts/work/1/s/src/azure-cli/azure/cli/__main__.py:16:1: E402 module level import not at top of file
/home/vsts/work/1/s/src/azure-cli/azure/cli/__main__.py:17:1: E402 module level import not at top of file
/home/vsts/work/1/s/src/azure-cli/azure/cli/__main__.py:18:1: E402 module level import not at top of file
/home/vsts/work/1/s/src/azure-cli/azure/cli/__main__.py:19:1: E402 module level import not at top of file
6     E402 module level import not at top of file

Thus, per-file-ignores won't work wither.

Since there is no support for file-level # flake8: noqa to disable certain errors (https://gitlab.com/pycqa/flake8/-/issues/399), temporarily comply to flake8's E402 rule to make the CI pass.

[jiasli]

I am not sure if this running time info should be moved to logger.debug. I am open to this.

[haroldrandom]

Would it better to calculate the execution time under certain situations such as debug mode, --verbose mode, etc.

But this modifcation doesn't bring any (significant) regression, So, LGTM.

[jiasli]

Would it better to calculate the execution time under certain situations such as debug mode, --verbose mode, etc.

But this modifcation doesn't bring any (significant) regression, So, LGTM.

The execution time is calculated regardless of the execution mode (verbose, debug). --verbose and --debug only decide if the execution time is printed.

Anyway, the overhead of calling timeit.default_timer() is pretty small. So we don't need to worry about it.