Add opt-in timing tests for the compare editor open path

CompareOpenPerformanceTest measures wall time from opening a compare
editor until the text is visible and until the diff is reported, for
several synthetic file sizes and change densities, plus a unified diff
scenario and a headless DocumentMerger microbenchmark. Each scenario is
repeated, warmup repetitions are discarded, and the median is printed
as a greppable PERF line and written to a CSV; only a generous sanity
bound is asserted, so shared CI timing noise cannot produce spurious
failures. RangeDifferencerTest is migrated off the retired
org.eclipse.test.performance database harness, which asserted nothing
without a baseline database, onto the same repeated-run median pattern
while keeping its correctness assertions. The suite is intentionally
not part of AllCompareTests; it runs on demand via PerformanceTestSuite.

Contributes to #2795

Author: vogella

team/tests/org.eclipse.compare.tests/META-INF/MANIFEST.MF

@@ -7,8 +7,9 @@ Require-Bundle: org.eclipse.compare,
  org.eclipse.jface.text,
  org.eclipse.jface,
  org.eclipse.core.runtime;bundle-version="[3.29.0,4.0.0)",
- org.eclipse.test.performance,
  org.eclipse.ui;bundle-version="3.208.0",
+ org.eclipse.ui.ide,
+ org.eclipse.ui.workbench.texteditor,
  org.eclipse.core.resources,
  org.eclipse.core.tests.resources,
  org.eclipse.core.tests.harness,