The expect_bounce_fail () helper assumed that messages could always
be sent. However in some cases zmq_send() blocks, due to there not
being any outgoing pipe. This changed in 77f5f7, where previously
there would be a pipe that kept trying to reconnect forever.

Solution: use a send timeout and check for EAGAIN if sending failed.

PLAIN: Small simplification

Add a send timeout in expect_bounce_fail()

Set an alarm(60) in setup_test_environment() when not on windows so test...

will not block too long if they go wrong.

failures don't block.

Don't reconnect on protocol errors

Fix bug in reporting protocol errors

Problem: artificial restriction on binary identities

Tell the session why the engine has stopped

Applications that use ZMQ_IDENTITY can be trapped by the artificial
restriction on not using a binary zero as first byte. It's specially
nasty on random generated identities, e.g. UUIDs, as the chance of a
binary zero is low, so it will pass 255 out of 256 times.

Solution: remove the restriction.

CURVE: Implement server-side ERROR handling

CURVE: Implement client-side ERROR handling

Prefix error-reason with length in ERROR command

NULL: Correct ERROR command format

Finish ERROR handling for PLAIN mechanism

PLAIN: Implement ERROR handling in client

test case for zmq_getsockopt/ZMQ_IDENTITY_FD

Simplify VS project configs and harden MSBUILD script

Update cmake, mingw32 and msvc builds

This became necessary after b9c27481