check itemgroup to exist and be instance of ModifiableTopLevelItemGroup
fix name .substring

Unclear whether 'function() {...}();' is syntactically valid (works in Chrome & HtmlUnit tests) or whether we need '(function() {...})();' as Rhino seems to demand and various other scripts use.

Continuing c9cf3b67, retain jenkinsRule but empty as a placeholder, and register its rules in the new way.
Registering them with increasing positions, since there seems to be some implicit order dependency involving TR.rowvg-start.

Not so visible an effect currently due to a bug fixed in 4ace69e (maven-hpi-plugin 1.87)
by which extensions of all plugins which are test deps of jenkins-test-harness were being loaded inadvertently.
Originally tried to define ${bundledPlugins} using gmaven-plugin as per [1] but this does not work,
so adopted the tack of just using a static glob *.hpi and interpreting that in core.
[1] http://stackoverflow.com/questions/7601835/why-cant-the-maven-deploy-plugin-resolve-my-custom-system-property

fix JENKINS-14772

Retain hudsonRules/jenkinsRules for now, as many scripts in core & plugins directly modify this hash.

Callers seem to expect that priority behaviors are applied before any behaviors in other selectors.

as get(String idOrFullName) suggest

so that getAll() don't invoke CanonicalIdResolvers for known IDs