Wrt additions; I agree, trailing comma is something I have seen, and consider specific problem to resolve. Empty values I am neutral about; but if it is needed, it can be handled.
As usual, my main concern is that there is no measurable overhead at parser level for standard use case – this because low-level parser is heavily profiled, optimized, and seemingly small changes can (alas!) throw JVM/Hotspot off. It usually just means that I need to be careful to measure before/after.
As to configurable handling: one possibility would be to add "error character handler". Question is what such a handler could do (skip character; parse it and more; materialize token(s)?), and how it could interface with parser internals. This could be a way to allow arbitrary extensions; not necessarily easily, but at least make it possible. I think that idea warrants a separate RFE, where we could brainstorm. I actually like the idea if it could be fleshed out; this would specifically allow extension outside of core parser.