Use a single squashed commit for downstream syncs

[jgraham]

In the past we had the ambition of running Gecko CI for each commit of a GitHub PR. However that didn't work out, and we ended up only starting the sync process after the PR had merged. However we're still applying the commits from the PR one at a time, rather than squashing and applying the whole diff like we do for landings. This increases the chance of merge conflicts, and makes it harder to deal with PRs that have a strange merge history. We should change to just squash for downstream syncs, and consider cherry-picking the squashed commit for the landing, rather than recreating it.