Customize boto3/botocore client instances to use the AWS CLI credential cache

[hannes-ucsc]

The code added in 4b1c051 and further improved in 0c91f99 works when the python interpreter is invoked from the same shell that _preauth was invoked from, but not when running/debugging code from within PyCharm (unless PyCharm was launched from a shell that already ran _preauth).

One solution would be to consolidate boto3 client instantiation in a central place and configure the client to use the AWS CLI credential cache just like _preauth function does. Care must be taken to not break, slow down or interfere with client allocation in other contexts, like Travis, Gitlab or AWS lambda.

Also see this PR for an previous effort to consolidate the client instantiation.

[jessebrennan]

A potential solution is to initialize boto3's botocore session to use a session that links to the CLI cache. This initialization should happen within envhook.py script so that it affects every Python process.