Implement retries in e2e tests in CI

[jpraynaud]

Why

We have noticed that the majority of flakiness cases in the CI is due to the Cardano devnet not starting correctly and during the era marker writes on chain and era marker writes on chain steps. These errors are due to the very fast epoch length and slot length parameters used and they mean that the whole Cardano devnet needs to be restarted.

What

Provide a retry mechanism to avoid having the CI fail when the Cardano devnet has not started correctly.

How

• Catch the Cardano devnet errors and make them return a specific error code by the end to end tester
• Implement retries in the CI and the nighly backward compatibility workflows:
  • with action https://github.com/marketplace/actions/retry-step
  • calibrate the number of retries to 2