Support AWS credentials override via environment variables

[zstumgoren]

It's quite possible that multiple collaborators on a data project will have multiple AWS profiles, and that the aws profile name for the shared resource do not match ~/.aws/credentials. For example, one user could have the shared creds under "default", while another user has them under "work".

While members could sidestep this issue by duplicating the credentials under a new, common name, it would be nice to support profile customizations by user.

We should update the data push and pull commands so that precedence is given to environment variables, if present when invoking aws s3 sync.

/cc @myersjustinc

[myersjustinc]

Here's the list of environment variables the AWS CLI supports.

[zstumgoren]

@myersjustinc Thanks for the link. The plugin uses the --profile flag by default, so we'll need to build in a mechanism at the datakit layer to check for the presence of environmental variable(s) and give those precedence. Open to suggestions on whether to check for the AWS_DEFAULT_PROFILE env var or the combo of AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY. Leaning toward the former since it simplifies implementation and reenforces usage of the ~/.aws hidden directory. Could also support both, though not sure if absence of the region env var will cause problems in the latter scenario.