Our cost model should be self-consistent, and the relative values for
the remote tuple and startup costs should reflect their real cost,
relative to costs of other operations like CPU tuple cost.
For example, now remote costs are set even lower than the parallel tuple
and startup cost. Contrary to that, their real world cost is going to be
an order of magnitude higher or more, because parallel tuples are sent
through shared memory, and remote tuples are sent over the network.
Increasing these costs leads to query plan improvements, e.g. we start
to favor the GROUP BY pushdown in some cases.
A follow-up for the review comments in the previous PR.
1. Create one backport PR per one source PR, even with multiple commits.
1. Add a comment to the source PR if we fail to backport it for some
reason.
1. Find the latest release branch
1. For each commit in main and not in release branch (compared by
title), find the corresponding PR.
1. If the PR fixes an issue labeled "bug", and neither the PR nor the
issue are labeled "no-backport", cherry-pick the commits from the PR
onto the release branch, and create a PR with these changes.
