Easy enough to code, but you're looking for a service that no e-mail provider I know of provides, but there are people who provide that type of service.
When you tell your e-mail provider (in this case, Google) to not mark a message as spam, you're effectively turning off their spam detection. If they're no longer detecting, then how could they mark. There are online e-mail inspection sites that could receive your message and give it a spam score, but then you'd still need a way for that score to be delivered to you without your mail carrier using their spam filter, or that score, to move it out of your inbox.
The most effective way I could imagine would be to let your carrier do their own spam filtering, and use a protocol that recognizes folders, so you can pull the messages that they've marked. I've already suggested this, but as you pointed out, you've rejected this option prior to even posting.
The only way I can see you using a 26 year old protocol (RFC 1081) that was designed only for downloading messages in a single mailbox, is to disable your online mail provider's detection so it's all in one box (you've already done this), and do 100% of your spam detection on your client. This is possible to have SMTP servers submit messages to online evaluators who will give you back a score, but it's still 100% on your client to determine if it's spam, Google's detectors have already been disabled by your process. SpamAssassin is one such evaluator.