Fixing namespace issues for variables

[Roland McGrath]

> re_syntax_options (bug 18442) may be a more straightforward case for this 
> fix, as a pure GNU extension.

Well, it's simpler only in the sense that there is no conformance issue to
think about.  It's probably also less sensitive in that the likelihood that
this identifier is in use in existing application code seems low off hand.
But I think the primary concern for all the cases is whether it's really OK
to effectively reserve an identifier for all purposes rather than just for
external linkage when that identifier was previously available for other uses.