I think the main problem with any such system is that a paper has a certain weight. You can be sure that it has a minimum value. (and estimate its value from the journal it's published in). Number of contribuons does not give an equally good indicator - eg the most active users on Wikipedia are all bots doing minor edits.
In addition, there is no consensus building like on github for individual projects which each in turn can be judged on popularity/use/... . Having a similar system for a science corpus of knowledge would even more encourage status quo thinking than the current system.
Apart from that, having a self-contained paper has value as that's a package you can judge as relevant or not to your work. It's fairly easy to use as the author is basically doing the curation. Having all knowledge of a field in one single Wikipedia like entity could end up being overwhelming for anyone but (and maybe even) existing experts.
Github has no barrier to entry, yet does not have a prestige issue. People get hired based on the content of their github account. I can't see why funding agencies couldn't use a similar system to track the academic contributions of researchers.
It's ironic that the rest of the world has moved on to connected, consistent experiences while academics are still stuck with papers. I've been thinking after I submit my thesis that I should find other people that are likewise frustrated and try to build a better system for academic collaboration and dissemination. Perhaps something more like Wikipedia than a journal.