Document how RenderNode properties form the transform matrix

 *     }

 * </pre>

 *

 * <p>A few of the properties may at first appear redundant, such as {@link #setElevation(float)}

 * and {@link #setTranslationZ(float)}. The reason for these duplicates are to allow for a

 * separation between static & transient usages. For example consider a button that raises from 2dp

 * to 8dp when pressed. To achieve that an application may decide to setElevation(2dip), and then

 * on press to animate setTranslationZ to 6dip. Combined this achieves the final desired 8dip

 * value, but the animation need only concern itself with animating the lift from press without

 * needing to know the initial starting value. {@link #setTranslationX(float)} and

 * {@link #setTranslationY(float)} are similarly provided for animation uses despite the functional

 * overlap with {@link #setPosition(Rect)}.

 *

 * <p>The RenderNode's transform matrix is computed at render time as follows:

 * First a setTranslate(getTranslationX(), getTranslationY()) is applied to a {@link Matrix}.

 * Second a preRotate(getRotationZ(), getPivotX(), getPivotY()) is applied to the matrix. And

 * finally a preScale(getScaleX(), getScaleY(), getPivotX(), getPivotY()) is applied. The current

 * canvas transform matrix, which is translated to the RenderNode's position,

 * is then multiplied by the RenderNode's transform matrix. Therefore there is no implicit

 * ordering in setting various RenderNode properties. That is to say that:

 *

 * <pre class="prettyprint">

 *     renderNode.setTranslationX(100);

 *     renderNode.setScaleX(100);

 * </pre>

 *

 * is equivalent to

 *

 * <pre class="prettyprint">

 *     renderNode.setScaleX(100);

 *     renderNode.setTranslationX(100);

 * </pre>

 *

 * <h3>Threading</h3>

 * <p>RenderNode may be created and used on any thread but they are not thread-safe. Only

 * a single thread may interact with a RenderNode at any given time. It is critical