History writes fail after restoring history indices onto a cluster that already has its own #2313

What is the bug?
Alerting writes alert, finding and comment history through write aliases (.opendistro-alerting-alert-history-write, .opensearch-alerting-finding-history-write, .opensearch-alerting-comments-history-write). If the cluster created its own history index before the snapshot's history indices are restored, the restore brings the alias back on the restored index too, because a restore keeps aliases. The alias then names two indices and neither is the write index, so every history write and rollover through it fails. Monitors keep running, so nothing shows the problem. This doesn't happen when the restore sets include_aliases: false, because the restored indices then come back without the alias.

What is the expected behavior?
After a restore, each history write alias names one index, and history writes and rollover keep working.