Add several angle bugs for end2end tests failing on NVIDIA

src/tests/gl_tests/GLSLTest.cpp

@@ -1511,6 +1511,7 @@ TEST_P(GLSLTest, MaxVaryingVec3ArrayAndOneFloatArray)
 TEST_P(GLSLTest, TwiceMaxVaryingVec2)
 {
     // TODO(geofflang): Figure out why this fails on NVIDIA's GLES driver
+    // (http://anglebug.com/3849)
     ANGLE_SKIP_TEST_IF(IsNVIDIA() && IsOpenGLES());
 
     // TODO(geofflang): Find out why this doesn't compile on Apple AMD OpenGL drivers

src/tests/gl_tests/ImageTest.cpp

@@ -1623,6 +1623,7 @@ TEST_P(ImageTest, MipLevels)
     // Driver returns OOM in read pixels, some internal error.
     ANGLE_SKIP_TEST_IF(IsOzone() && IsOpenGLES());
     // Also fails on NVIDIA Shield TV bot.
+    // http://anglebug.com/3850
     ANGLE_SKIP_TEST_IF(IsNVIDIAShield() && IsOpenGLES());
     // On Vulkan, the clear operation in the loop is optimized with a render pass loadOp=Clear.  On
     // Linux/Intel, that operation is mistakenly clearing the rest of the mips to 0.

src/tests/gl_tests/MipmapTest.cpp

@@ -1093,6 +1093,7 @@ TEST_P(MipmapTestES3, GenerateMipmapCubeBaseLevel)
 
     // Observed incorrect rendering on NVIDIA, level zero seems to be incorrectly affected by
     // GenerateMipmap.
+    // http://anglebug.com/3851
     ANGLE_SKIP_TEST_IF(IsNVIDIA() && IsOpenGL());
 
     // Draw using level 0. It should still be blue.