Automate S3 backups of release attachments

[jwnimmer-tri]

Is your feature request related to a problem? Please describe.

We currently have a "push s3" feature of the push_release script that we run manually after every release:

drake/tools/release_engineering/dev/push_release.py

Lines 347 to 348 in 54a304e

	if options.push_s3:
	_push_s3(state)

This is extra work that would be better automated.

Describe the solution you'd like

• Amend the continuous-mirror-to-s3 job to do this for us.

• Remove the code that does this from push_release.

(Possibly this just means factoring out the code that does this from push_release into a separate program, designed for CI to call.)

[tyler-yankee]

For reference, the current program is drake/tools/workspace/mirror_to_s3.py. It's currently designed to work with the drake-mirror bucket, only, and gets its GitHub bits from read_repository_metadata.

drake/tools/workspace/metadata.py

Line 12 in 54a304e

def read_repository_metadata():

CI invokes this script directly out of bazel-out/. (The "intended usage" atop the script says to use bazel-bin/, so if we want to be exact, perhaps that could use updating as part of this ticket also?)

My first impression is that trying to combine the scripts logically doesn't make sense, but I could see modifying CI to simply call both on a mirror-to-s3 job, and modifying the existing push_release.py to follow a similar logical workflow (i.e., check if expected exists in S3, if not then grab from GitHub and mirror).

[jwnimmer-tri]

My first impression is that trying to combine the scripts logically doesn't make sense, ...

Yes, I had the same feeling when I looked, too.

[jwnimmer-tri]

Also for clarity -- the automation only needs to check/backup the current release at the time it's run. No need for it to go back in time and check old releases.