Bugfix/sm degraded performance

[matt-land]

This version addresse the following:

• migrate python 2.7 to 3.7
• enable API retries instead of immediate failure when SecretManager when performance is degraded
• change the lambda timeout to 300 seconds to allow for retries
• add terraform module outputs of the secret name & arn to allow sharing the secret with other terraform modules
• add a randomizer to the secret name, so that multiple terraform apply/destroy/apply do not enconter the "failed to create secret" errors where the secret already exists, but is deleted
• scope the lambda assume role policy document to allow the lambda access to one secret only, not all secrets in the account