TODO for parsifal
-----------------

* Memory allocation could be "heaps better", string interning, smart heap to
  avoid memory fragmentation etc...
* More profiler runs and optimizations...
* Maybe internal widechar support (wchar_t) by conditional compilation? maybe not.
* There's lot of room for improvement in optimization department.

FOR VALIDATION
--------------

* Checking of non-deterministic content models
* Testing/better handling of duplicate attribute declarations (for bsearch)
* DTD caching (already designed). Can propably be implemented as separate
  suspended parser (using XMLFLAG_USE_SIMPLEPULL) which holds DTD properties.
