Fix flaky test, add retry to testcloud

[bajertom]

Pull Request Checklist

• implement the feature
• extend the test coverage

Fixes #4557

First test mentioned above sometimes fail due to the timeout being too fast and cuts off right before /plans/multi/plan/one gets through, so I've added another expect phrase that comes after that line and increased timeout to 3 seconds.

Second test imho failed due to messed up download of the qcow2 image.

        progress: downloading...
Unknown download size.
        effective hardware: {}

The unindented "Unknown download size." error seems odd. I've checked successful runs of that test in other PRs and "Unknown download size" does not appear there. Another line says:

qemu-img: /tmp/tmp.9726GhCN02/testcloud/instances/tmt-001-RuKJKCkN/tmt-001-RuKJKCkN-local.qcow2: Image is not in qcow2 format

So I think that something did get downloaded but the file was corrupted.

I wrapped the responsible code part into a function prepare_image and added a validation of the qcow2. If that fails, the downloaded file will be removed and prepare_image will be retried.

Hard to fix flaky test when they're flaky and can't be reproduced at will 😐

Co-generated by Claude Code.

[gemini-code-assist]

Code Review

The pull request refactors the testcloud image preparation logic to include validation and retry mechanisms. It introduces new constants for image preparation retry attempts and interval, imports a retry utility, and wraps the image preparation and validation steps in a retry loop. A new validation step uses qemu-img info to verify the integrity of downloaded qcow2 images, removing corrupt ones and raising a ProvisionError if validation fails. Review comments suggest improving the descriptive comment for the new retry constants and enhancing the clarity and readability of the ProvisionError message when a corrupt image is encountered.

[tcornell-bus]

If any of these exceptions occur, the function will get retried by rety(). Do they have a chance of succeeding if retried? If not, only 15 seconds are wasted, so it doesn't seem like a huge deal. Just wondering if it was intended that these would also cause retries.