Prevent starting null read watcher

[dkropachev]

Fix: #503

This issue happens when connection is opened and closed right away.
_loop_will_run calls conn.close on it, which removes self._read_watcher from the connection and since connection stays in self._new_conns on next iteration it will pick it up and try to start self._read_watcher on it.
Solution is to remove closed connection from self._new_conns and check if self._read_watcher is not None before starting it.

Two fixes for follwing problem:

Exception ignored in: <bound method LibevLoop._loop_will_run of <cassandra.io.libevreactor.LibevLoop object at 0x7343276afc10>>
Traceback (most recent call last):
  File "/123/cassandra/io/libevreactor.py", line 197, in _loop_will_run
    conn._read_watcher.start()
AttributeError: 'NoneType' object has no attribute 'start'

1. Check if self._read_watcher is not None before starting it
2. Remove connection from self._new_conns when it is destroyed

Pre-review checklist

• I have split my patch into logically separate commits.
• All commit messages clearly explain what they change and why.
• I added relevant tests for new features and bug fixes.
• All commits compile, pass static checks and pass test.
• PR description sums up the changes and reasons why they should be introduced.
• I have provided docstrings for the public items that I want to introduce.
• I have adjusted the documentation in ./docs/source/.
• I added appropriate Fixes: annotations to PR description.