retried_test_reports: send "reports" instead of alerts

If alerts are configured in Long tests, then all exceptions thrown at run-time are sent to the configured slack channel for alerts.

The previous implementation of retried_test_report.ml configured alerts and used alerts to send the flake report.

Turns out that tezos-sw-metrics fail quite a lot due to rate limiting.

To avoid spamming the clients of the reports with said failures, this MR implements a separate "reporting" mechanism, uses this for retried_test_report and unconfigures the alerts.