Give each redshift client their own boto session

[jweibel22]

resolves #2756

Description

Since the boto session is not thread-safe, using the default session in a multi-threaded scenario will result in concurrency errors.
This change ensures that a fresh session is instantiated before creating the client. This is similar to what is happening in the else branch. Thus, this change should not cause any unexpected issues (e.g. exceeding service quotas)

Checklist

• I have signed the CLA
• I have run this code in development and it appears to resolve the stated issue
• This PR includes tests, or tests are not required/relevant for this PR
• I have updated the CHANGELOG.md and added information about my change to the "dbt next" section.

[jtcohen6]

@jweibel22 Nice! I'd love to get this in for v0.18.1. Does this change work for you locally when you run with multiple threads?

Also, could you:

• pull changes from dev/0.18.1 (which should include fixes to the failing Redshift integration tests)
• add a changelog entry under v0.18.1

[cla-bot]

Thank you for your pull request and welcome to our community. We could not parse the GitHub identity of the following contributors: Github Build Bot.
This is most likely caused by a git client misconfiguration; please make sure to:

1. check if your git client is configured with an email to sign commits git config --list | grep email
2. If not, set it up using git config --global user.email [email protected]
3. Make sure that the git commit email is configured in your GitHub account settings, see https:/settings/emails

[jweibel22]

@jweibel22 Nice! I'd love to get this in for v0.18.1. Does this change work for you locally when you run with multiple threads?

Yes, I have verified that is does resolve the issue when I run dbt compile locally.

[beckjake]

This looks very nice!

Could you add a test to the redshift unit tests in tests/unit/test_redshift_adapter.py that makes sure we call boto3.Session().client() instead of boto3.client() directly in both the iam_profile cases (none and not none)? I would hate to regress on this.

[beckjake]

Thanks! This looks great to me

[jtcohen6]

Thanks for this @jweibel22! I'll try to resolve the changelog conflict and get this in.