RemoteDockerImage should fail immediately on fatal errors, ability to specify a custom pull retry timeout

Right now RemoteDockerImage will try to re-pull container image on error until the default timeout of 2 minutes has been reached.

This is a reasonable thing to do to work around various I/O and network issues, but we should still allow user to specify a custom timeout and update the code to fail immediately on fatal errors (e.g. credential helper not available or invalid credentials configured).

Background, Context

If you utilize Docker image from some private registry (e.g. ECR, DockerHub, Quay, etc.) and credential helper for that registry is not available or credentials configured correctly, that could will try to retry up to 2 minutes before failing.

Exception similar to this one will be visible in the logs:

2021-02-23T12:34:35.050+0100 [DEBUG] [TestEventLogger]          at java.base/java.lang.Thread.run(Thread.java:834)
2021-02-23T12:34:35.051+0100 [DEBUG] [TestEventLogger]     [Test worker] WARN 🐳 [fooo.dkr.ecr.us-east-1.amazonaws.com/ingestion-pipeline:local-redis-cluster] - Retrying pull for image: foo.dkr.ecr.us-east-1.amazonaws.com/bar:foo (112s remaining)
2021-02-23T12:34:35.530+0100 [DEBUG] [TestEventLogger]     [docker-java-stream--1357434991] ERROR com.github.dockerjava.api.async.ResultCallbackTemplate - Error during callback
2021-02-23T12:34:35.530+0100 [DEBUG] [TestEventLogger]     com.github.dockerjava.api.exception.InternalServerErrorException: Status 500: {"message":"Get \"https://foo.dkr.ecr.us-east-1.amazonaws.com/v2/bar/foo\": no basic auth credentials"}

Proposed Change

I proposed updating that code to fail immediately if no basic auth credentials message is returned (https://github.com/Kami/testcontainers-java/blob/master/core/src/main/java/org/testcontainers/images/RemoteDockerImage.java#L87). I know that checking the error message is not the most robust thing, but it's probably the best we can do here.

Technically, Docker Java client could also be updated to throw some more specific exception in that case, but that would require more work.

Start with core/src/main/java/org/testcontainers/images/RemoteDockerImage.java, particularly the retry logic around the referenced line 87, and trace how Docker pull errors are surfaced. Done means callers can provide a custom pull retry timeout and fatal credential errors such as “no basic auth credentials” fail without waiting for the default timeout.