Add profiling information for inference example

[Xreki]

All inference examples can be profiled with this PR.
The profile result of recognize_digits_conv on my server:

$ ./paddle/fluid/inference/tests/book/test_inference_recognize_digits_conv --dirname=/home/liuyiqun01/PaddlePaddle/Paddle/python/paddle/fluid/tests/book/recognize_digits_conv.inference.model --repeat=100
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from inference
[ RUN      ] inference.recognize_digits
WARNING: Logging before InitGoogleLogging() is written to STDERR
I0307 08:28:55.945766 22596 test_inference_recognize_digits.cc:29] FLAGS_dirname: /home/liuyiqun01/PaddlePaddle/Paddle/python/paddle/fluid/tests/book/recognize_digits_conv.inference.model
I0307 08:28:58.983011 22596 test_inference_recognize_digits.cc:51] --- CPU Runs: is_combined=0 ---

------------------------->     Profiling Report     <-------------------------

Place: CPU
Time unit: ms
Sorted by event first end time in descending order in the same thread

Event                       Calls       Total       Min.        Max.        Ave.        
thread0::init_program       1           1.0292      1.0292      1.0292      1.0292      
thread0::conv2d             200         25.4355     0.068962    0.538576    0.127177    
thread0::elementwise_add    300         3.86118     0.00526     0.030406    0.0128706   
thread0::relu               200         1.48954     0.005296    0.024998    0.0074477   
thread0::pool2d             200         4.79081     0.014388    0.059552    0.0239541   
thread0::batch_norm         100         1.59581     0.014327    0.03309     0.0159581   
thread0::mul                100         1.59436     0.01552     0.01875     0.0159436   
thread0::softmax            100         0.925062    0.008602    0.035497    0.00925062  
thread0::run_inference      100         72.8846     0.654562    1.08013     0.728846    

I0307 08:28:59.059149 22596 test_inference_recognize_digits.cc:54] 1, 10
I0307 08:28:59.059178 22596 test_inference_recognize_digits.cc:62] --- GPU Runs: is_combined=0 ---

------------------------->     Profiling Report     <-------------------------

Place: CUDA
Time unit: ms
Sorted by event first end time in descending order in the same thread

Event                       Calls       Total       Min.        Max.        Ave.        
thread0::init_program       1           10.8645     10.8645     10.8645     10.8645     
thread0::conv2d             200         20.242      0.071392    0.1608      0.10121     
thread0::elementwise_add    300         7.44586     0.007392    0.067872    0.0248195   
thread0::relu               200         3.20122     0.005824    0.034112    0.0160061   
thread0::pool2d             200         8.21853     0.037696    0.059424    0.0410926   
thread0::batch_norm         100         6.44045     0.060096    0.095584    0.0644045   
thread0::mul                100         3.73216     0.034784    0.061344    0.0373216   
thread0::softmax            100         5.14694     0.047328    0.076448    0.0514694   
thread0::run_inference      100         100.706     0.955872    1.29904     1.00706     

I0307 08:28:59.189551 22596 test_inference_recognize_digits.cc:65] 1, 10
I0307 08:28:59.189587 22596 test_inference_recognize_digits.cc:51] --- CPU Runs: is_combined=1 ---

------------------------->     Profiling Report     <-------------------------

Place: CPU
Time unit: ms
Sorted by event first end time in descending order in the same thread

Event                       Calls       Total       Min.        Max.        Ave.        
thread0::init_program       1           0.315289    0.315289    0.315289    0.315289    
thread0::conv2d             200         23.926      0.068842    0.313521    0.11963     
thread0::elementwise_add    300         3.85191     0.005079    0.027506    0.0128397   
thread0::relu               200         1.41673     0.005313    0.012125    0.00708363  
thread0::pool2d             200         4.66253     0.014217    0.038472    0.0233126   
thread0::batch_norm         100         1.55707     0.01443     0.022271    0.0155707   
thread0::mul                100         1.60407     0.01551     0.037741    0.0160407   
thread0::softmax            100         0.902546    0.008676    0.01315     0.00902546  
thread0::run_inference      100         70.3325     0.679056    0.863227    0.703325    

I0307 08:28:59.262215 22596 test_inference_recognize_digits.cc:54] 1, 10
I0307 08:28:59.262234 22596 test_inference_recognize_digits.cc:62] --- GPU Runs: is_combined=1 ---

------------------------->     Profiling Report     <-------------------------

Place: CUDA
Time unit: ms
Sorted by event first end time in descending order in the same thread

Event                       Calls       Total       Min.        Max.        Ave.        
thread0::init_program       1           0.65248     0.65248     0.65248     0.65248     
thread0::conv2d             200         20.3624     0.070976    0.177568    0.101812    
thread0::elementwise_add    300         7.47549     0.00736     0.047872    0.0249183   
thread0::relu               200         3.10365     0.005856    0.02992     0.0155182   
thread0::pool2d             200         8.17149     0.037984    0.061472    0.0408574   
thread0::batch_norm         100         6.37984     0.061152    0.07664     0.0637984   
thread0::mul                100         3.70675     0.03504     0.039968    0.0370675   
thread0::softmax            100         4.96397     0.045664    0.05664     0.0496397   
thread0::run_inference      100         99.6111     0.951648    1.08646     0.996111    

I0307 08:28:59.380663 22596 test_inference_recognize_digits.cc:65] 1, 10
[       OK ] inference.recognize_digits (3435 ms)
[----------] 1 test from inference (3435 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test case ran. (3435 ms total)
[  PASSED  ] 1 test.

[sidgoyal78]

Thanks for the PR. I think this looks good to me.