Fix known issues with PDF building

[lastoCHka42]

• Firs point of the table of content always goes up-leveled. As
  the result, the Overview page replaces with 3 top-level points.
• In the Connectors section, we have some extra pages we shouldn't
  have

Follows up #3549
Related to #3548

[lastoCHka42]

We don't have extra pages in the Connectors section, that's just about how Sphinx parses the connectors.rst.
For the JSON build, it creates full page connectors and then adds pages from toctree. As a result, we see paragraphs Protocol, Packet example, Interpreting function return values as a part of Connectors starting page, so Sphinx don't include them as the ToC items.
In the PDF and singlehtml versions, Sphinx first builds pages from the toctree section (C, Go, Java, Python, Community supported connectors). As a result, paragraphs after the toctree section are no more part of the Connectors starting page, so they are independent ToC items.

After discussion with @veod32, we've decided not to fix it.