a) the default properties for the code point make it not split a word it
appears in into two different words in any break mode we have. Which is what we
want from a CH_TXTATR_INWORD

b) unicode TR#20 gives for the interlinear annotation anchor: "What to do if
detected: In a proxy context or browser context, remove U+FFF9", so when we
need to strip it from text to run that text through e.g. the spellchecker or
word counting then there's a solid precedent for stripping it

In addition I *do* want the footnote placeholder to break the word it appears
in, that gives the desired wordcount and cursor travelling behaviour

The BREAKWORD and other *random* selection of CH_TXTATR are still odd choices,
and there's way too many of them.

Change-Id: I930ff8ff806af448829bc1a1ae6cb92053e9a284