- Fixes a regression in core where the display name clear on copy was triggering a save
- More than one way to do this, could also have used the marker interface approach
  This route seems slightly less fragile, though people could still add ItemListeners
  with order == -Double.MAX_VALUE which would then introduce intdeterminism.
  A marker interface would remove that indeterminism as the onCopyComplete method would
  be only called on the Job as the last method... but it could be hard to
  ensure that all ItemGroupMixin's respect the calling of onCopyComplete contract
  hence this approach seems better to me for that reason

- The caller is no longer responsible as the listener is actually invoked from the method

Use DescribableList to handle the copy-on-write semantics correctly. The vector class just doesn't cut it, and we've been setting a new value to this field, which will violates all sorts of the concurrent programming practice.

This change has the nice side effect of removing {{class="vector"}} from the persisted XML. A test is added to make sure we can still read back such an XML.

Revisiting the fix. Avoid parsing the entire text by only looking at the tail portion.

This will hopefully reduce the file locking problems on Windows.

see
https://github.com/jenkinsci/jenkins/commit/2ffece9d737f4a77ed4059c777948299daf22771#commitcomment-3526425

Do not use anymore shaded version of wagon-http as it we can control commons-logging version

- if the user is anonymous, he'd go to HudsonAUthenticationEntryPoint, not here.
- even if there is some code path that comes here, there's no guarnatee that SecurityRealm in charge wants a login form.

This allows HudsonAuthenticationEntryPoint to access which permission was needed

... before making any modification.

The original version is acegi-security 1.0.7

Node monitoring should run in parallel to reduce the total round-trip
time in large instances.

Conflicts:
	test/src/main/java/org/jvnet/hudson/test/HudsonTestCase.java

Integrated the fix toward 1.521.