avoids -Wall failures comparing int to uint

tests that default LINGER (-1) ensures message delivery.

Note: it fails for large messages ( > 1MB, depending on the system).

Change name of icanhasall to subscribe_to_all

icanhasall is cute (for now), but the effect of the variable is clear
only after tracking down its origin reading the commit. This change is
intended to make it easier for people to have some intuition about its
effect from its name.

Disable failing tests on windows.

Fixed silly port number.

Fixed markup so documentation can be built.

get to clean compile under vs2012

Added some missing (but existing in branch) files to libzmq11 project needed
to compile as static library by VS2012 (at least), and fixed some warnings.
One warning remains in raw_decoder.cpp, which I'm leaving for now because I'm
not sure how project owners would like to fix it.

Add the Socket Identity info to the ZAP Messages

Add ZMQ_CONFLATE socket option

This change adds the socket identity infomartion from the socket to the
zap frames.  In doing this the ZAP is able preform different operations
based on different sockets.  This is not compaitable with the current ZAP
RFC, but that can be updated.  As the ZAP rfc is currently draft for I
did not change the version number.

Tests also modified and passing.

Minor doc & assert fixes

Optimized zmq::msg_t for constant messages

Attempt to fix disconnect not respecting linger

Port tests to windows and add to cmake build.

Looks like linger is honoured properly, but shutting down the session
causes the pipe termination to come from that side - because the local
pipe then shuts down right away it seems to trigger a terminated on the
other end instead of waiting. This way we trigger the termination from
the local end and then terminate the session.

Add zmq_send_const for constant data

ZMQ_CONFLATE option is passed to pipepair() which creates a usual
ypipe_t or ypipe_conflate_t and plugs it into pipe_t under a common
abstract base.

Add simple double buffer implementation tailored to handle msg_t,
i.e invoke msg_t::close instead of destructor and so on.
Seems to me mutex is good enough at this point.

Required for memmove(), memcpy()