OpenSSL timeout fix: (backport)

[azat]

Since timeout was already set via setsockopt, we call wait_io_or_timeout() with a very small timeout (5ms) to get a more precise errno, which is used by OpenSSL's error function.

Note (azat): initial patch (mariadb-corporation#250) was wrong:

Georg Richter:

> Timeouts were already set via setsockopt() function before. Since
> the socket already timeout, it will wait andother read/write timeout
> period.  Instead of, we should get wait_io_or_timeout with a timeout of
> a few milliseconds to get the correct errno.

And got reverted in (mariadb-corporation#253).

Upstream: mariadb-corporation@a6fd09f

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[azat]

Actually I have couple of questions about upstream patch, let's wait until I will get answers