Make sure that fpm_event_add calls inside a timer callback work by
unregistering the event from the queue before invoking its callback.

The read timeout in tester.inc is increased because the added test
needs two seconds (one for SIGTERM, one for SIGKILL) until the
reload succeeds, so we should wait longer than that for a response.

AC_FPM_POLL defines HAVE_POLL symbol which is already checked by
configure.ac and function poll in AC_CHECK_FUNCTIONS.

Closes GH-4449

Closes GH-4451.

This reverts commit e0eca262852dba1a78afcde64a49126c81fead1a.

free_filename is used by the wincache extension, restore this
field for PHP 7.4.

The php_stream_read() and php_stream_write() functions now return
an ssize_t value, with negative results indicating failure. Functions
like fread() and fwrite() will return false in that case.

As a special case, EWOULDBLOCK and EAGAIN on non-blocking streams
should not be regarded as error conditions, and be reported as
successful zero-length reads/writes instead. The handling of EINTR
remains unclear and is internally inconsistent (e.g. some code-paths
will automatically retry on EINTR, while some won't).

I'm landing this now to make sure the stream wrapper ops API changes
make it into 7.4 -- however, if the user-facing changes turn out to
be problematic we have the option of clamping negative returns to
zero in php_stream_read() and php_stream_write() to restore the
old behavior in a relatively non-intrusive manner.

Seems to be very unreliable in CI.

4 seems to be enough to reliably reproduce the issue. Let's see
if this works better in CI.