todo keywords can be set after the todos themselves. this means we must scan the
'#+TODO' settings first in the first pass, then re-scan the document to parse
the elements. to scan the '#+TODO' settings we can simply use a regex since the
settings must be put at the beginning of lines.
* FOO foo
* BAR bar
* BAZ baz
* TODO blah
#+TODO: FOO | BAR BAZ