Minimize new refs produced by validateWithRef

• For tests using validateWithRef, only the preprocessed stdout is written out on failure. Standard preprocessors would filter out the summary tables, which is why the counter tables, TTree and histogram summaries are extracted and written out at the end. In this way irrelevant lines would not be part of the ref files, thereby reducing diffs and conflicts.
• Fix a bug in the counter extraction where any component (tool) with a dot in the name would not be considered in the comparison.
• Catch exceptions in the validators and put the traceback in a validator_exception field.
• Expose the counter parsing with a standalone extract_counters.
• Add a preprocessor GroupMessages that groups multi-line messages.
• Add a preprocessor BlockSkipper that improves upon the BlockSkipper from Gaudi (can be moved there).

goes together with Moore!339 (merged)

GaudiConf/python/GaudiConf/QMTest/LHCbExclusions.py

@@ -9,7 +9,8 @@
 # or submit itself to any jurisdiction.                                       #
 ###############################################################################
 import os
-from GaudiTesting.BaseTest import normalizeExamples, RegexpReplacer, LineSkipper, SortGroupOfLines, BlockSkipper
+from GaudiTesting.BaseTest import (normalizeExamples, RegexpReplacer,
+                                   LineSkipper, SortGroupOfLines, BlockSkipper)
 
 gitCondDBFixes = (RegexpReplacer(
     when='Detector description database:', orig='conddb:/', repl='git:/') +
@@ -92,3 +93,9 @@ preprocessor = (
 from DDDB.Configuration import GIT_CONDDBS
 if GIT_CONDDBS:
     preprocessor = preprocessor + gitCondDBFixes
+
+# Exclude counters with
+counter_preprocessor = LineSkipper([
+    ' | "Delta Memory/MB" ',
+    ' | "Total Memory/MB" ',
+])