Additional tests for string arrays and collections

[mbellade]

Following up on #585 with additional tests as suggested. I've added tests for POJOs with a string array or with a list, both are behaving as expected. I also added a test for a simple POJO with just plain string properties to be extra sure nothing breaks.

Regarding this comment:

Ok, this explicit comment makes me nervous: it specifically spells out that when asking from nextTextValue(), empty element should not be coerced into null

I was unsure about that comment too, but it looks like that particular token switch case XmlTokenStream.XML_END_ELEMENT only occurs for null string values (when xsi:nil writing and processing is enabled). The XmlTokenStream.XML_TEXT case is the one that gets triggered in all other cases as well as for empty strings.

[cowtowncoder]

@mbladel First of all, big thank you for these tests!

And yes, wrt null induction, I was wondering about the exact same thing. Conversion of xsi:nil is not well commented in code but I think your analysis is spot on.
This assessment makes me much more confident in the solution.