IBM tests

Thanks, Sandy, for committing these tests. They will prove a great resource.

First results from Saxon: about 5 tests caused software crashes, in all 
cases bad error recovery after reporting errors in invalid schemas. 
After fixing those, I've got 32 discrepancies in terms of expected 
results, out of 800 or so tests. That seems pretty good for the first 
exposure of new tests to a new product-under-test, and speaks well of 
the tests, the implementation, and the spec.

I look forward to examining the discrepancies!

Michael Kay
Saxonica