This patch implements the text-transform property. It correctly handles every 
English-language nuance that I have been able to think of. It adds fields to 
each FOText object that point to previous & next FOText objects within the same 
Block, so that the text in related FOText objects can be considered for needed 
context. It then transforms the text as part of the FOText Constructor. Some of 
the fields and methods added may be useful for other parts of FOP as well. This 
code assumes that words do not span multiple blocks.

Caveat: I cannot think of any, but if there are cases (now or in the future)
where FOText objects do not have ancestor Block objects, text-transform will 
not be able to consider other FOText objects for context. This can be fixed by 
storing the Block-like ancestor instead.

Caveat: The definition of "word" for purposes of this property is ambiguous. It 
may need to be tweaked over time, especially for scripts other than Latin. I 
have tried to build an infrastructure that will allow such modifications to be 
pretty easily implemented.

Victor Mote (vic@outfitr.com)