Fix: Apparently using the len output of a failed match somewhere

[xparq]

Or just not initializing it to 0 for non-consuming operators.
-> That was it, indeed, in the new _DEF op. (probably called from a _SEQ loop)!

Now in match():

#ifdef NDEBUG
		len = 0; //! This doesn't help (it's even bad, for false sense of sec., and
		         //! easily masking bugs with that benign-looking 0 in the output),
		         //! as 0 is a valid output, which should still be ignored -- as any
		         //! others! -- if match() returned false!
			 //! OK, disabling it in debug builds for better diagnostics, but
			 //! enabling in release mode for some cushioning!...
#else
		len = (unsigned)-666; // And indeed, this *did* crash! :) (e.g. #29)
#endif