Yeah, I'm also surprised to see people so upset about the idea of the feature mostly for the argument that it doesn't presently exist. As for the compile-time-extension, my main concern with that as a free software project is that we try to play nice with distro packaging and requiring users to compile their own sqlite would be setting them up for a bit of pain. But I suspect it doesn't matter: if the feature was coded with the aim of pushing it upstream, I'm suspecting there's a good chance upstream would accept it if of good quality.
Code is unlikely to be accepted upstream because it very rarely is (assuming you mean by the SQLite team). They are very picky about code contributions and almost always implement themselves. One part is legal - there has to be very clear lineage/authoring to the code, patent clearances etc.
The second part is that the actual code to implement any particular functionality or even a bug fix is relatively trivial. There is far more effort and far more lines of code for testing: http://www.sqlite.org/testing.html - and implementing code to enable the rigourous testing usually results in a different structure.
Then again, I'm not upstream. :)