Remove blind settle delay from code mining test fixtures

CodeMiningTest.setUp() and CodeMiningLineHeaderAnnotationTest.setUp()
both ended with DisplayHelper.sleep(display, 1000), placed directly
after a waitForCondition() that had already established that the text
widget is visible. DisplayHelper.sleep() is waitForCondition() with the
condition hardcoded to false, so it always burns the full second.

CodeMiningTest installs its provider on an empty document, and the
provider skips blank lines, so no code minings are produced during
setUp and there is nothing to settle for.

Measured on a full CI run, CodeMiningTest is 22.6s over 19 tests with a
median of 1.04s against a minimum of 0.19s, and
CodeMiningLineHeaderAnnotationTest is 4.0s over 4 tests with minimum
and median both 1.01s.

Author: vogella

tests/org.eclipse.jface.text.tests/src/org/eclipse/jface/text/tests/codemining/CodeMiningLineHeaderAnnotationTest.java

@@ -70,7 +70,6 @@ protected boolean condition() {
 				return fViewer.getTextWidget().isVisible();
 			}
 		}.waitForCondition(display, 3000));
-		DisplayHelper.sleep(textWidget.getDisplay(), 1000);
 	}
 
 	@AfterEach

tests/org.eclipse.jface.text.tests/src/org/eclipse/jface/text/tests/codemining/CodeMiningTest.java

@@ -132,7 +132,6 @@ protected boolean condition() {
 				return fViewer.getTextWidget().isVisible();
 			}
 		}.waitForCondition(display, 3000));
-		DisplayHelper.sleep(textWidget.getDisplay(), 1000);
 	}
 
 	@AfterEach