Slightly refactor new rabbitmqadmin code

[tianon]

This is a follow-up to #780 -- I'm happy to discuss any of this further (and to be explicitly clear, I'm very open to changes!) ❤️

• align on "bashbrew architectures" for representing the architecture strings (matching the Architectures: lines in our downstream library/rabbitmq file; this makes it easier to filter that list later if we so choose)
• avoid multi-stage build, especially to avoid the empty file copying/weirdness
• we always need the rabbitmqadmin data, so we should query it before the loop (and pre-process it since the data is the same for every single iteration of the loop)
• store version in the rabbitmqadmin block (our automation uses this for making nicer commit messages like "Update 4.0 to rabbitmqadmin 2.17.1" instead of just "Update 4.0" which is what the current code will lead to)
• use dpkg --print-architecture instead of uname -m for architecture detection (the former is "userspace architecture" detection, which more closely matches what we need than the latter's "kernel architecture" detection)

[lukebakken]

👍

[yosifkit]

waiting to remove the Python-based version until at least the next patch release if not the next minor/major?

So as to not break users, should we wait for version bumps before we push the alpine-based images that no longer have a rabbitmqadmin or should we go ahead with the changes? i.e. should we add back the python-based version in the alpine images until the next patch of each release series?