Further optimize rendering methods

[BenCheung0422]

This change performs well together with #523 and #524. As a minor change, there is no longer an error popup when there is any explicitly or implicitly incompatible images regarding their sizes. Instead, a warning message is rendered in the terminal. Due to the change to the compatibility of the rendering of the on-screen keyboard, the textures of the special keys are updated and saved as assets instead of being constructed on demand in runtime. It is mainly because the visibility of Screen#pixels is decreased, There is no longer a light screen constructed with Screen. Instead, an object representing the overlay is present in the screen object. Other minor changes:

• Screen is constructed with a parameter of BufferedImage instead of using the initialization method #init(int[]) to initialize the pixel array.
• Screen now stores BufferedImage for the ability to render through Graphics2D on the image object.
• Most of the rendering methods except sprite rendering (compatibility problem) are replaced by using Graphics2D provided method.
• Methods to render an arbitrary size of area of MinicraftImage are added, because of the change to the on-screen keyboard.
• All methods to render now add queues instead of being rendered instantaneously. All the renderings are executed only when Screen#flush() is called and all processes before the latest screen clean rendering are eliminated.
• The drop in FPS on the world surface is significantly diminished.

Some methods may be still using array looping to render.

As a result, the CPU run time has been improved. The below comparison is tested based on my personal computer. Comparing with the current implementation (with #523):

• Before entering a world: improved by around 50%
• In world: improved by around 30% Comparing with #524 added (comparing without this pull request):
• Before entering a world: improved by around 60%
• In world: no obvious change

[Litorom]

Fix Conflicts and I’ll merge