X10 2.6.0kobeu+stl+team patch

See merge request x10kobeu/x10kobeu!1

Conflicts resolved:
	x10.runtime/src-x10/x10/util/Team.x10

Conflicts:
	x10.runtime/src-x10/x10/util/Team.x10

	modified:   x10.runtime/src-cpp/x10/xrx/Thread.h
	modified:   x10.runtime/src-java/x10/core/Thread.java
	modified:   x10.runtime/src-x10/x10/xrx/Thread.x10
	modified:   x10.runtime/src-x10/x10/xrx/Worker.x10

	modified:   x10.runtime/src-cpp/x10/xrx/Thread.h
	modified:   x10.runtime/src-java/x10/core/Thread.java
	modified:   x10.runtime/src-x10/x10/xrx/Thread.x10
	modified:   x10.runtime/src-x10/x10/xrx/Worker.x10

the default affiity of each thread is configurable by using X10RT_DEFAULT_CPUMAP environment variable (i.e. each thread will be bound to processor 0-3 where X10RT_DEFAULT_CPUMAP="0,1,2,3" : * only support comma separeted format)
modified:   x10.runtime/src-cpp/x10/xrx/Thread.cc
modified:   x10.runtime/src-cpp/x10/xrx/Thread.h

barrier before native communication methods are disabled when the environment variable "X10_TEAM_WITHOUT_BARRIER" is set to "1"

the default affiity of each thread is configurable by using X10RT_DEFAULT_CPUMAP environment variable (i.e. each thread will be bound to processor 0-3 where X10RT_DEFAULT_CPUMAP="0,1,2,3" : * only support comma separeted format)
modified:   x10.runtime/src-cpp/x10/xrx/Thread.cc
modified:   x10.runtime/src-cpp/x10/xrx/Thread.h

resolve #1

Closes #1

See merge request !3

Revert "add patch to evacuate activities in queues that will be parked"

See merge request !2

This reverts commit 2ac9ee40

Revert "move and change evacuation method"

See merge request !1

This reverts commit 242d52ef

barrier before native communication methods are disabled when the environment variable "X10_TEAM_WITHOUT_BARRIER" is set to "1"

The code was failing to properly propagate the parent finish
to the full-blown HC finish.  As a result, any full-blown HC
finish that was created when an optimized HC finish 'went global'
did not have a parent.

implement notifyActivityCreatedAndTerminated via a single coalesced operation on the resilient store.

avoid promoting to full-blown remote finish when notifyRCC is called; all that is actually needed is disabling the call to join in waitForFinish.

been created in put/get implementations to avoid scheduling
bugs in waitForFinish.  Will need to further optimize nRCC in the
resilient finish implementations in subsequent commit to avoid
losing performance on gets...