This makes it possible to avoid looping for a long time while working
with a fast enough peer when data are added to the socket buffer faster
than we are able to read and process them (ticket #1431).  This is
basically what we already do on FreeBSD with kqueue, where information
about the number of bytes in the socket buffer is returned by
the kevent() call.

With other event methods rev->available is now set to -1 when the socket
is ready for reading.  Later in ngx_recv() and ngx_recv_chain(), if
full buffer is received, real number of bytes in the socket buffer is
retrieved using ioctl(FIONREAD).  Reading more than this number of bytes
ensures that even with edge-triggered event methods the event will be
triggered again, so it is safe to stop processing of the socket and
switch to other connections.

Using ioctl(FIONREAD) only after reading a full buffer is an optimization.
With this approach we only call ioctl(FIONREAD) when there are at least
two recv()/readv() calls.

Normally, the epoll module calls the read and write handlers depending
on whether EPOLLIN and EPOLLOUT are reported by epoll_wait().  No error
processing is done in the module, the handlers are expected to get an
error when doing I/O.

If an error event is reported without EPOLLIN and EPOLLOUT, the module
set both EPOLLIN and EPOLLOUT to ensure the error event is handled at
least in one active handler.

This works well unless the error is delivered along with only one of
EPOLLIN or EPOLLOUT, and the corresponding handler does not do any I/O.
For example, it happened when getting EPOLLERR|EPOLLOUT from
epoll_wait() upon receiving "ICMP port unreachable" while proxying UDP.
As the write handler had nothing to send it was not able to detect and
log an error, and did not switch to the next upstream.

The fix is to unconditionally set EPOLLIN and EPOLLOUT in case of an
error event.  In the aforementioned case, this causes the read handler
to be called which does recv() and detects an error.

In addition to the epoll module, analogous changes were made in
devpoll/eventport/poll.

Avoid POLLREMOVE and itimerspec redefinition.

In theory, this can provide a bit better distribution of latencies.

Also it simplifies the code, since ngx_queue_t is now used instead
of custom implementation.

It's mostly dead code.  And the idea of thread support for this task has
been deprecated.