* ip_resolver_t: adding virtual descructor to silence C++-warning -Wnon-virtual-dtor

* adding my relicense

News out of date, formatting not clean

Solution: run make clang-format-diff

Solution: update it

Solution: add it

* add option to stop trying to reconnect on ECONNREFUSED

Fix unused-variable warning in perf/proxy_thr.cpp

Add relicense grant by Laszlo Boszormenyi (GCS)

Android build helper improvement + fix for prefix dir

Solution: Add usage and default value to README

Solution: Migrate the properties and the selection for the correct env
and abi to the helper. There's no need for the user to specify them.

Solution: Use pwd only a default value

Improve android build handling

Solution: Re-factor them to look like usual tables

Solution: Modify the build scripts so the user can specify the platform
for which to build, e.g. `./build.sh arm`. This approach originally
significantly reduces the parameters which have to be set before running
the script.

Further the build process is documented in a README now.

Add grant by ninoles

Received via email, message-id:

ed4de883-640b-bf7e-ac56-911e117b649e@tzone.org

problem: zeromq doesn't has a thread-safe peer to peer socket

Solution: a new socket type, called PEER. Very similar to SERVER, but can only connect to other PEERs. Also a new zmq_connect_peer method, that connect and return a routing-id in thread-safe and atomic operation

problem: maximum size of group doesn't match the RFC maximum size

Implement ZMTP 3.1 subscribe/cancel via commands

websocket problems

Solution: return after the fallback

fixes #3809

Grant from KLNikita

Received via email, message-id:
81ce91c8-df40-13c2-7abf-6f648d0b123f@elyzion.net

solution: if msg is not shared or constant, mask the message in place

Grant by asafkahlon

Received via email, message-id:

CALFsTFVZLgujRE3V_30G9V9b35kE566K5cWi0vKbNkZGSHrszA@mail.gmail.com

Solution: bump minor version number in the engine as all 3.1 features
are now implemented

Solution: if all peers of a socket are >= 3.1 use sub/cancel commands
instead of the old 0/1 messages.
For backward compatibility, move the handling of 0/1 or sub/cancel
command strings to the encoders, so that the right thing can be done
depending on the protocol version.
Do not set the command flag until the encoder, so that we can handle
the inproc case (which skips the encoder).

support compiling with IBM XL 16.1 w/ clang front end