Added fallback mechanism for specific socket binding on Windows platform with IPv6 enabled

Solution: honor it

* Doing so was zmq_poller_poll to always return 0.

This should restore full compatibility with earlier zmq_poll behavior.

It complicates things a little bit, as collisions must be detected, and when collisions are found:

- event masks must be merged
- pollitems, events arrays are no longer co-ordered

Reverts the recent zmq_proxy patch to workaround the lack of repeat-item support in zmq_poll that is now fixed.

not -1, ETIMEDOUT like zmq_poller

Return value is the number of events found. This also propagates to the return value of zmq_poller_wait_all.

zmq_poller_wait was only returning events on the first-registered socket.

zmq_poller doesn't allow a socket to appear twice

variable-size array allocation is not allowed by some compilers

enables zmq_poll on threadsafe sockets only supported in zmq_poller (radio, dish, etc.)

previous patch updated only the branch for ZMQ_POLL_BASED_ON_POLL, causing failed builds on Windows.

clang from Xcode 8 has these already defined, causing duplicate macro warnings

Solution: use only Libs.private to avoid breaking application builds.
Even though Requires.private are supposed to be parsed only if
pkg-config is called with --static, the --cflags parameter is enough
to trigger the parsing, causing build failures for applications that
do not (and should not) depend on libzmq's dependencies.

Solution: add dependencies, if necessary, to the .private Libs and
Requires field of the pkgconfig file at build time.
This way pkg-config --static --libs libzmq will correctly print
dependencies if they were used to build the static libzmq.a library.

instead of allocating a new, identical array and copying the data.

This is only safe while zmq_poller_event_t and zmq::socket_poller_t::event_t are the same struct,
which they presumably will remain.

avoids unnecessary heap allocations, races on the number of items

Solution: zmq_poller_wait_all signals all events

allows signaling multiple events with one call to zmq_poller_wait_all
rather than emitting only one event.

this prepares for zmq_poll being based on zmq_poller,
which requires events for all sockets rather than just one.

* fix bugs of the pollset

1. extend 'fd_table' when fd_ is greater or equal than the size of 'fd_table';
2. delete specific fd from pollset before reset pollin or pollout according the description of AIX document

* fix bugs of the pollset

edit error. remove extra spaces and paste fault

* fix bugs of pollset

remove character '-' at the end line.

fd_entries (std::vector) can reallocate underlying storage
which will render reference invalid.

Solution: add a zmq_assert to check if the ephemeral sockets created
to drain the queue of pending inproc connecting sockets was allocated
successfully.

Problem: zmq_ctx_term asserts with connect-before-bind and sockets with identity over inproc transport

Solution: check if the connecting inproc socket has been closed
before trying to send the identity.
Otherwise the pipe will be in waiting_for_delimiter state causing
writes to fail and the connect to assert when the context is being
torn down and the pending inproc connects are resolved.
Add test case that covers this behaviour.

Problem: Pub socket crashes on client disconnect.  family_entry.fd_entries [i] is modified in zmq::select_t::rm_fd()
Solution: do not copy family_entry.fd_entries [i]

add a new poller named pollset which will get benefit of performance in AIX platform.

Solution: allow for '[' character when doing the basic sanity check
on the TCP endpoint.
Also add unit tests for both IPv4 and IPv6 source;dest format.

Solution: remove them

Solution: replace tabes with spaces

Solution: Use only lower case for header file name.

We can find "wincrypt.h" by "WinCrypt.h" on Windows because Windows uses
case insensitive file system. But we can't find "wincrypt.h" by
"WinCrypt.h" on Linux Because Linux uses case sensitive file system.

Solution: Add a scope that scopegurad unlocks the mutex
 before the object instance deletes itself and wants to unlock
a locked mutex

Solution: add a lock guard that takes optional a mutex and use it

The gssapi has some helper functions gssalloc_malloc()/gssalloc_free()
which on windows doesn't call malloc()/free(). Instead these are
wrappers around HeapAlloc() and HeapFree(). To complicate matters
gssapi doesn't export these helper functions, so you're left using
the allocation method of your choice.

See Here:
https://github.com/krb5/krb5/blob/89683d1f135765e91041f3a239af865b11aaf86b/src/lib/gssapi/generic/gssapi_alloc.h

The zmq gssapi implementation is calling malloc and then calling
gss_release_buffer() to free the memory. gss_release_buffer uses
gssalloc_free() to free this buffer which on windows calls HeapFree()
instead of free(). This causes an access violation on windows.

Convert manual (locking and unlocking) mutexes to scoped mutexes for the case of unlocking mutex even if the protected operation throws an exception (#2071)

Linux provides accept4(2) which will return a socket with FD_CLOEXEC set
when called with the SOCK_CLOEXEC flag. So call this when available and
fall back to fcntl(..., FD_CLOEXEC) if not.

getifaddrs() can fail transiently with ECONNREFUSED on Linux.
This has been observed with Linux 3.10 when multiple processes
call zmq::tcp_address_t::resolve_nic_name() simultaneously.

Before asserting in this case, make 10 attempts, with exponential
backoff, given by (1 msec * 2^i), where i is the attempt number.

Fixes #2051