- fix iovec init problem introduced in prev recvmmsg patch.
  Implies small revert of some args and test changes.
  It also pushes de-exposes struct iovec again, which is good.
- unrelated, irrel TBB Writer micro-patch (sorry)

Changes are for APERTIF (and to fix COBALT...).

- rework recvmmsg() wrapper, so we can reuse it for APERTIF. Main purpose is to allow verification of incoming datagram sizes in real-time mode from recvmmsg().
- fix small bug wrt input statistics in non-real-time mode
- fix small bug wrt setReadBufferSize() (unused): THROW_SYSCALL() instead of perror()
- minor simplification / clean ups, e.g. merge/unify packet size check into validatePacket()

Disadvantage of the patch is that struct iovec is no longer contained in our recvmmsg() wrapper.
All tests pass on cbt009.

Also:
- Move deadline check from createStream() into SocketStream().
- Get rid of unused code in an old attempt by me to refactor createStream().
- Fix inverted check > 0 vs < 0 on one of the 2 usleep()s to back-off connect() in SocketStream().
- Move EndOfStreamException and TimeOutException out of their classes (and fix up users).
- Replace readdir() by readdir_r(). Probably irrel since only retval checked vs 0.
- Clarify ~FileDescriptorBasedStream() somewhat, only delay cancellation if actually calling close().
- Constify a few local vars in SocketStream.cc.

Reason: In APERTIF, we need to receive UDP frames from the FPGA correlator.
We can live with LOFAR Common and Stream, but prefer not to have CoInterface just for createStream().
(All other network communication will be done via QPID messaging middleware.)

Build tested only.

Task #6639: cherry-pick merge from LOFAR-Release-2_5_0-PVSS-event-overload-Task6639. All can go to the trunk except for the .log_prop to DEBUG changes (r30126)

Task #5921: Reverted all changes on this branch. I'm going to solve this issue in a more pragmatic and less intrusive way.

Task #5946: Moved broadcast mechanism to (simulated) MPI_Ibcast; Do not continue reading commands on Exception

Task #5946: Stations now send an EOS packet when done. The pipelines stop if all stations sent an EOS, allowing the senders to decide when sending has completed, instead of the Parset dictating the number of blocks.

Task #5841: Removed parset.settings.rawStartTime, by regenerating it. Code from RSPTimeStamp to convert seconds (double) to a string has been moved to TimeFuncs.h because to accomplish this.

Task #5844: change 'Aborted' msg into 'Stopped' on valid/expected EINTR. Aborted scares as it is printed on abort().

Task #5830: revert most KVpair changes. The extra overloads clash on either 32 or 64 bit. Adding overloads will also fail, as they cannot be mapped to PVSS data types in the PVSSGateway (default case there is to try string). So overloads -> template won't work either. Agreed with Ruud to also remove the vector<int> support (already not mapped). Then to solve the original data type problem, adapt a few logged data types from uint or ulong to int. While at it, adapt a few to bool from string ("0" or "1").

Task #5781: Attempted reintegration of task branch into trunk (error has been reported: svn: E155023: Can't set properties on'/data1/schoenma/LOFAR_trunk/src/LOFAR/RTCP/Interface': invalid status for updating properties.

Task #5830: fix up some data points for PVSS: obsId as string, remove array notation from InputProc stream data points

Task #5883: mpi is a global variable no more, and (typically) needs to be declared in each program instead.

Task #5883: Replace MPI_Init/Rank/Size/Finalise with wrapper class, reducing the code base and making sure that MPI_Finalize is *really* called at the end, and not before MPI_Free_mem in particular

Task #5832 (in branch of Task #5830): add PVSS datapoint support for InputProc (as part of rtcp / GPUProc).

- Moved OMPThread.h to CoInterface
- Added tOMPThread
- Added CoInterface/TimeFuncs.h for struct timespec functions
- Added OMPThreadSet class to register/kill groups of OpenMP threads
- Use OMPThreadSet to be able to kill slow write() calls in both Pipelines

  Also add script to set the 'current' symlink. The changes are mostly in error checking.
  The 3 added .sh scripts are not yet tested in its entirety, only parts.
  They have been derived from working scripts, but need to be retested. This is best on the cbt009 after hostname redirection (incl cbt001-008 unavail).
  Also, they may need extra functionality once tested:
    restrict to cbt001-008 or cbt009 
    do not upload/download to/from NEXUS
  Move install location of a few scripts from bin/ to sbin/.

  Most changes are renaming. 'Station' idents are unclear wrt expect values "CS001" vs "CS001HBA0".
    Parset.cc has a clear "translation" point. Added 2 TODOs, as renaming these two functions is too invasive for this task.
  Add option to restrict cobalt nodes to those that receive streams from ant fields in the obs.
    Default off, because as JD mentioned, this may exclude nodes that are needed for computing only.

Task #5555: Compute delays in caller's thread, simplifying the code. Caller is a light-weight thread already (GPUProc/Station/StationInput::StationMetaData)