The xml() function was being passed a flag for pretty printing the XML in word/document.xml. This was causing some parsers (not Word itself) to break when they didn't expect whitepsace. It was also causing some files to be prettified and others to be ugly. This ensures consistent prettification.
Rework the way attributes are stored in ImportedXmlComponent to match elsewhere (required allowing for a null xmlKeys in the XmlAttributeComponent interface).
Rework the way paragraphs get added to the end of table cells if needed.
The goal in both reworks is to not mess around with the objects output from `prepForXml` if we can avoid it.
Made the output of RunProperties, ParagraphProperties, TableCellProperties, TableRowProperties, and TableProperties all optional based on whether they contain any attributes or children. Changed code in PageBorders, TableCellMargin, and TableCellBorders that implemented this same thing by overriding `prepForXml` so that it uses the new XmlComponent subclass instead.
Removed commented out code that attempted to fix-up XML output and make proper empty elements.
Fixed all affected tests.
Turn off `no-null-keyword` in the linter as we need to use null to signal to the `xml` library to create an empty element with no attributes (`undefined` will not work in its place).
Fixes#306
