Docs: update tooling and outdated/broken content

* Update docs build config
  * Specify Sphinx and theme versions in docs/requirements.txt
    (tox had been using latest; RTD had pinned older versions)
  * Update docs builds for Python 3 and recent Sphinx
  * Remove obsolete sphinx-rtd-theme patches
  * Add .readthedocs.yml
  * Move some JS to async (now that Sphinx supports that)
  * Fix version-alert.js for async loading
* Remove docs versionadded/changed info older than v3.0
* Fix broken GitHub issue links in changelog (never worked?)
* Add setup long_description_content_type
* Drop poorly maintained AUTHORS.txt 
  (just point to GitHub contributors page)

docs/inbound.rst

@@ -3,8 +3,6 @@
 Receiving mail
 ==============
 
-.. versionadded:: 1.3
-
 For ESPs that support receiving inbound email, Anymail offers normalized handling
 of inbound events.
 
@@ -168,7 +166,7 @@ Normalized inbound message
         :class:`~anymail.utils.EmailAddress` object, which makes it easier to access
         the parsed address fields:
 
-        .. code-block:: python
+        .. code-block:: pycon
 
             >>> str(message.from_email)  # the fully-formatted address
             '"Dr. Justin Customer, CPA" <jcustomer@example.com>'
@@ -367,11 +365,6 @@ have these methods:
         The errors param is as in :meth:`~bytes.decode`. The default "replace" substitutes the
         Unicode "replacement character" for any illegal characters in the text.
 
-        .. versionchanged:: 2.1
-
-            Changed to use attachment's declared charset by default,
-            and added errors option defaulting to replace.
-
     .. method:: get_content_bytes()
 
         Returns the raw content of the attachment as bytes. (This will automatically decode