These fixes seemed desirable while surveying L1 implementations

1. Changed conditionals using implementationAttribute 
ignoringElementContentWhitespace to equivalent test on node count.

Many tests had different expectations if element content whitespace was 
ignored.  However depending on an implementation attribute requires a reliable 
means of externally detecting whether the implementation preserves or discards 
element content whitespace.  However, the tests can easily check the number of 
child nodes to determine what actually happened.

2. Added assertNotNulls to prevent NullPointerExceptions

Many tests are failed by implementations returning null when a non-null value 
was expected.  If unchecked, these result in a NullPointerException that does 
not indicate the point of failure unless you look up the reported line number 
in the source code.  AssertNotNull's were added at common failure points to 
provide an assertion ID to simplify interpretation.

3. Better assertion names

Many assertion names were something like the test name + "Assert1".  Having 
names that are indicative of the failure makes it much easier to interpret 
test failures, especially if similar checks have the same name.

4. Removed redundant hasFeature("XML") preconditions.

All the non-hc_ tests require an XML implementation since staff.xml doesn't 
have an HTML equivalent.  Adding hasFeature("XML") would cause these tests to 
be skipped by DOM4J which returns false for some reason.

5. Changed preserve entity reference conditionals to observed node type.

Similar to issue 1.

6. Fixed metadata types

Several of the EE tests metadata contained the title of the base test.

7. SVG related fixes

Several tests did not adapt to SVG data files, for example, expecting "staff" 
as the document element name instead of "svg" when content type was 
image/svg+xml.