Minor fixes and config updates

[JWCook]

This fixes the following minor issues I found while testing this out:

• When specifying author in the fields list (rst) or frontmatter (markdown), it gets passed to feedgen as a string instead of as a dict:

  Traceback (most recent call last):
  File "sphinx/events.py", line 97, in emit
  results.append(listener.handler(self.app, *args))
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "sphinxfeed.py", line 102, in create_feed_item
  item.author(author)
  File "feedgen/entry.py", line 363, in author
  self.__atom_author += ensure_format(author,
                        ^^^^^^^^^^^^^^^^^^^^^
  File "feedgen/util.py", line 61, in ensure_format
  raise ValueError('Invalid data (value is no dictionary)')

• When using the sphinx-tags extension, its tags directive gets picked up as page metadata, as a list instead of a string:

  Traceback (most recent call last):
    File "sphinx/events.py", line 97, in emit
      results.append(listener.handler(self.app, *args))
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    File "sphinxfeed.py", line 107, in create_feed_item
      tags = tags.split()
             ^^^^^^^^^^
  AttributeError: 'list' object has no attribute 'split'

As well as a few other small changes I needed:

• Use a sphinx config option feed_use_atom to generate an Atom feed instead of RSS
• Use Sphinx config instead of rstgen (to avoid adding as a dependency) to get use_dirhtml config value
• Log when a page is skipped due to a date in the future (using sphinx.util.logging like you suggested)
• Updated the test output for Sphinx 7.4 and docutils 0.21

I did notice that in the HTML output, I no longer see the "generator" meta tag that docutils usually adds, but that isn't related to the RSS feed.