Add benchmarking tool.

Takes two qbs repo states, runs callgrind and massif on different phases
of the build process for each of them and informs the user of the
relative performance of these operations.
I suggest running this tool before committing anything that could
conceivably impact performance. Note that we have similar regression
detection functionality on our build machines, but it's better to spot
such problems before pushing a change.
Example output (for commits 04df1532c4 and f53d724eec building qbs itself):
========== Performance data for Resolving ==========
    Old instruction count: 1931242543
    New instruction count: 1457438375
    Relative change: -25 %
    Old peak memory usage: 1710721 Bytes
    New peak memory usage: 1509879 Bytes
    Relative change: -12 %
========== Performance data for Rule Execution ==========
    Old instruction count: 2924308821
    New instruction count: 2890322188
    Relative change: -2 %
    Old peak memory usage: 2520740 Bytes
    New peak memory usage: 2488058 Bytes
    Relative change: -2 %
========== Performance data for Null Build ==========
    Old instruction count: 572213104
    New instruction count: 517813513
    Relative change: -10 %
    Old peak memory usage: 1964455 Bytes
    New peak memory usage: 1774279 Bytes
    Relative change: -10 %

Change-Id: I1f05c647d204b6cacf3539c8ecbf13633b757cf7
Reviewed-by: Joerg Bornemann <joerg.bornemann@theqtcompany.com>

1 files changed, 1 insertions, 1 deletions

diff --git a/tests/tests.pro b/tests/tests.pro
index 8a102073b..2ababcb8c 100644
--- a/tests/tests.pro
+++ b/tests/tests.pro
@@ -1,2 +1,2 @@
 TEMPLATE = subdirs
-SUBDIRS = auto fuzzy-test
+SUBDIRS = auto fuzzy-test benchmarker