docs/sphinx: fix extra stuff in TOC after freeform QMP sections

Freeform sections with titles are currently generating a TOC entry for
the first paragraph in the section after the header, which is not what
we want.

(Easiest to observe directly in the QMP reference manual's
"Introduction" section.)

When freeform sections are parsed, we create both a section header *and*
an empty, title-less section. This causes some problems with sphinx's
post-parse tree transforms, see also 2664f317 - this is a similar issue:
Sphinx doesn't like section-less titles and it also doesn't like
title-less sections.

Modify qapidoc.py to parse text directly into the preceding section
title as child nodes, eliminating the section duplication. This removes
the extra text from the TOC.

Only very, very lightly tested: "it looks right at a glance" :tm:. I am
still in the process of rewriting qapidoc, so I didn't give it much
deeper thought.

Reported-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Message-ID: <20240822204803.1649762-1-jsnow@redhat.com>

docs/sphinx/qapidoc.py

@@ -388,6 +388,7 @@ def _start_new_heading(self, heading, level):
         self._active_headings[level - 1] += snode
         self._active_headings = self._active_headings[:level]
         self._active_headings.append(snode)
+        return snode
 
     def _add_node_to_current_heading(self, node):
         """Add the node to whatever the current active heading is"""
@@ -417,13 +418,11 @@ def freeform(self, doc):
             # the first line of the block)
             (heading, _, text) = text.partition('\n')
             (leader, _, heading) = heading.partition(' ')
-            self._start_new_heading(heading, len(leader))
+            node = self._start_new_heading(heading, len(leader))
             if text == '':
                 return
 
-        node = self._make_section(None)
         self._parse_text_into_node(text, node)
-        self._add_node_to_current_heading(node)
         self._cur_doc = None
 
     def _parse_text_into_node(self, doctext, node):