[LOGGING_4864] Consider timeout errors and add logging debug statements

[jsubirat]

Description

A customer notified us that the retry mechanism was not properly considering TimeoutErrors, and even opened a ticket suggesting a solution. After inspecting the issue, we found that what was really performing the timeout was the ClientSession established for the whole lambda execution, which was set to tool low (3 seconds, I wonder if it was meant to be 30 instead). This PR fixes this issue by:

• Defining a timeout of 3 seconds for individual POST requests
• Defining a session timeout which is the sum of the expected individual POST requests (assuming the worst case of having a timeout in each of them), the backoffs between them and an additional time buffer of 1 second, to account for the processing time needed to create the payloads/requests.
• The customer also complained that it was rather difficult to debug what was failing of our lambda, so we added a debug mode that enables log.debug statements when retries are performed, as well as providing more details on the errors taking place.

Related ticket(s)

https://newrelic.atlassian.net/browse/LOGGING-4864
https://newrelic.atlassian.net/browse/LOGGING-4700

Tests

• The PR includes new unit tests for the added/modified functionalities.
• Manual tests have been performed in AWS lambda, using a fake API introducing controlled timeouts.

[MattWhelan]

Looks good. If you could just run black to reformat your changes to get past the lint error, we can merge this.

[MattWhelan]

@jsubirat Please amend your commits, sign them, and force-push.