CLOCK_MONOTONIC_FAST, like CLOCK_MONOTONIC_COARSE, has low accuracy. It
shows noticeable timing variation for short intervals, which is visible
in metrics like $upstream_response_time for fast upstream responses.
This change complements the work started in commit f29d7ade5.
In addition to the reasons described in f29d7ade5, the performance of
CLOCK_MONOTONIC is good enough on modern hardware when using a TSC
timecounter. This is especially true when it is accessed through a
shared page, as implemented in FreeBSD 10.0 (see git commits
869fd80fd449 and aea810386d8e for details).

Co-authored-by: Sergey Kandaurov <pluknet@nginx.com>

When configured, it enables Multipath TCP support on a listen socket.
As of now it works on Linux starting with Linux 5.6 and glibc 2.32,
where it is enabled with an IPPROTO_MPTCP socket(2) protocol.

To avoid EADDRINUSE errors in bind() and listen() when transitioning
between sockets with different protocols, SO_REUSEPORT is set on both
sockets.  See f7f1607bf for potential implications.

Based on previous work by Maxime Dourov and Anthony Doeraene.

It is believed to be harmless, see a similar change 077a890a76ff.

Reported-by: geeknik <geeknik@protonmail.ch>

For instance, the last_buf flag is used in the http proxy module when
creating HTTP/2 requests to indicate the output is closed.  The flag
is inherited in ngx_output_chain() to a destination buffer when reading
the buffered request body.  Then it is used in the proxy output filter
to mark the last HTTP/2 DATA frame with END_STREAM.

The problem happens when reusing the destination buffer, such as to
re-read the buffered request body on next upstream, because this buffer
may contain a dirty last_buf value, which breaks sending HTTP/2 request
body in multiple output filter calls.

The flush and last_in_chain flags are cleared for consistency.