[client] egl: add framework for multiple scaling algorithms

This commit fixes the issues with the meaning of useNearest being flipped
by removing the variable and use enumerations.

We define an enumeration EGL_DesktopScaleType to express the type of scaling
needed to be performed: no scaling, upscaling, or downscaling. This is
updated when either the frame size or the viewport size changes.
Previously, we only updated the useNearest when the frame size changes.

The desktop shader can now support an enumeration of scaling algorithms,
although only nearest and linear are currently implemented.

Like before, nearest is used when not scaling or upscaling, and linear is
used when downscaling.

client/renderers/EGL/shader/desktop_rgb.frag

@@ -1,11 +1,14 @@
 #version 300 es
 
+#define LG_SCALE_NEAREST 0
+#define LG_SCALE_LINEAR  1
+
 in  highp vec2 uv;
 out highp vec4 color;
 
 uniform sampler2D sampler1;
 
-uniform       int   nearest;
+uniform       int   scaleAlgo;
 uniform highp vec2  size;
 uniform       int   rotate;
 
@@ -36,10 +39,16 @@ void main()
     ruv.y =  uv.x;
   }
 
-  if(nearest == 1)
-    color = texture(sampler1, ruv);
-  else
-    color = texelFetch(sampler1, ivec2(ruv * size), 0);
+  switch (scaleAlgo)
+  {
+    case LG_SCALE_NEAREST:
+      color = texelFetch(sampler1, ivec2(ruv * size), 0);
+      break;
+
+    case LG_SCALE_LINEAR:
+      color = texture(sampler1, ruv);
+      break;
+  }
 
   if (cbMode > 0)
   {