misuse of cvPreprocessIndexArray + FIX (Bug #517)

thanks! Since cvPreprocessIndexArray is used by the other stat. models, not only the decision trees, it's safer to fix the root cause of the problem - the mentioned condition in the cvPreprocessIndexArray. It has been commented off in r3879.

r3879 breaks cascade training. opencv_traincascade and opencv_haartraining don't work with these two lines commented (see r3879). If I run current trunk version, application finished with exception and the output is following:

h5. TRAINING 0-stage

<BEGIN
POS count : consumed   3000 : 3000
NEG count : acceptanceRatio    1500 : 1
Precalculation time: 2
+----+---------+---------+
|  N |    HR   |    FA   |
+----+---------+---------+
|   1|        1|        1|
+----+---------+---------+
|   2|        1|        1|
+----+---------+---------+
....
+----+---------+---------+
|  99|        1|        1|
+----+---------+---------+
| 100|        1|        1|
+----+---------+---------+
END>
[[OpenCV]] Error: Assertion failed (tempNode->left) in write, file /home/kir/work/ros_opencv_trunk/opencv/modules/traincascade/boost.cpp, line 616
terminate called after throwing an instance of 'cv::Exception'
  what():  /home/kir/work/ros_opencv_trunk/opencv/modules/traincascade/boost.cpp:616: error: (-215) tempNode->left in function write

./scripts/trainCascade.sh: line 17: 23355 Aborted                 $cmd

But if I uncomment these two lines, everything is OK:

h5. TRAINING 0-stage

<BEGIN
POS count : consumed   3000 : 3000
NEG count : acceptanceRatio    1500 : 1
Precalculation time: 3
+----+---------+---------+
|  N |    HR   |    FA   |
+----+---------+---------+
|   1|        1|        1|
+----+---------+---------+
|   2| 0.996667| 0.389333|
+----+---------+---------+
END>

Fixed in 4102.