Fixes not receiving unsubscription messages in XPUB socket with
ZMQ_XPUB_VERBOSE and using a XSUB-XPUB proxy in front.

This adds two modifications:

- It adds a new flag, ZMQ_XPUB_VERBOSE_UNSUBSCRIBE, to enable verbose
  unsubscription messages, necessary when using a XSUB/XPUB proxy.

- It adds a boolean switch to zmq::mtrie_t::rm () to control if the
  callback is invoked every time or only in the last removal. Necessary
  when a pipe is terminated and the verbose mode for unsubscriptions is
  enabled.

Previously, AddressSanitizer flagged leaks when running
tests/test_many_sockets.cpp.

I traced memory leaks found using AddressSanitizer down to these.

Fixed wrong message::init arguments.

LIBZMQ-195 allow explicitly setting sndbuf and rcvbuf to 0 (see https://support.microsoft.com/en-us/kb/201213)

- Replaced stdlib.h with cstdlib
- Made single-argument constructors explicit

A memcpy is eliminated when receiving data on a ZMQ_STREAM socket. Instead
of receiving into a static buffer and then copying the data into the
buffer malloced in msg_t::init_size, the raw_decoder allocates the memory
for together with the reference-counter and creates a msg_t object
on top of that memory. This saves the memcpy operation.

For small messages, data is still copied and the receive buffer is reused.

Set the ZMQ_HEARTBEAT_TIMEOUT to default to the value of
ZMQ_HEARTBEAT_IVL if it's not explicitly set.
Change the units of ZMQ_HEARTBEAT_TTL to milliseconds in the API
and round down to the nearest decisecond so that all the options
are using the same units.
Make the maximum heartbeat TTL match the spec (6553 seconds)

This commit adds ZMTP connection heartbeats described in
http://rfc.zeromq.org/spec:37/ZMTP.

The shared reference count was not shared but copied. msg_t cannot
store the refcnt itsef but has to store a pointer to an externally
allocated (shared) refcnter. The changes to lmsg are reverted to
use content_t again. Howver, this introduces an allocation in v2_decoder
when creating the message which can be avoided. When allocating the reception
buffer, space is allocated for the maximum number of reference counts
(8192 / max_vsm_size = 8192/64 = 128 zmq:atomic_counter objects). This
increases the buffer by 128*sizeof(atomic_counter) = 128*4 = 512 bytes only.
When creating a message, the refcnt member is set to the address of one of the
pre-allocated atomic_counter_t objects. To do so, a new msg_t type zcmsg
is introduced because msg::copy must discriminate between the message types
when releasing memory.

zero-copy msg_t::init cannot be used when the message exceeds either
the buffer end or the last received byte. To detect this, the buffer
is now resized to the numnber of received bytes.

See #1220 recent comments. This change is untested yet seems
safe.

Fixes #1431.

Construct messages from a reference-counted buffer allocated once
per receive instead of copying the data.

With a msg_t size of 64 bytes, it becomes possible to embedd the content_t's members
struct for large messages directly in the msg_t. This saves the dynamic allocation
of content_t obejcts when using msg_t::init_data.

content_t contains a zmq::atomic_counter_t object which is not a POD in C++98
and thus it cannot be used as a member of the union u. To bypass this, C++11
is used which has relaxed rules for POD and atomic_counter is a C++11-POD. An
alternative would have been to make atomic_counter a classical POD by removing
constructors and all private member functions, i.e. have a struct and free functions
to manipulate it.

A new msg_t::init function is added which decides to either to copy the data for size<64 bytes
or use msg_t::init_data to do zero-copy initialization.

Of course people still "can" distributed the sources under the
LGPLv3. However we provide COPYING.LESSER with additional grants.

Solution: specify these grants in the header of each source file.

Fix another degradation, CPU maxes out when POLLOUT is set because poll exits on POLLOUT and doesn't wait for POLLIN.