[onert] Reallocated tensor when size becomes larger

[hseok-oh]

This commit updates dynamic allocator usage on basic tensor to reallocate when shape is changed to larger size.
It will improve performance and reduce memory usage when tensor size is lesser by dynamic shape on inference than prepare phase.

ONE-DCO-1.0-Signed-off-by: Hyeongseok Oh [email protected]

[hseok-oh]

Before this PR with #13654

$ ./Product/out/bin/onert_run -r 1 --shape_prepare [0,[4,299,299,3]] --shape_run [0,[3,299,299,3]] -m /home/nfs/model/inception_v3/inception_v3.tflite
Model Filename /home/nfs/model/inception_v3/inception_v3.tflite
===================================
MODEL_LOAD   takes 38.136 ms
PREPARE      takes 64.846 ms
EXECUTE      takes 127.666 ms
- MEAN     :  127.666 ms
- MAX      :  127.666 ms
- MIN      :  127.666 ms
- GEOMEAN  :  127.666 ms
===================================
RSS
- MODEL_LOAD   takes 108516 kb
- PREPARE      takes 143860 kb
- EXECUTE      takes 177556 kb
- PEAK         takes 177556 kb
===================================
HWM
- MODEL_LOAD   takes 110244 kb
- PREPARE      takes 145764 kb
- EXECUTE      takes 177712 kb
- PEAK         takes 177712 kb
===================================
PSS
- MODEL_LOAD   takes 195033 kb
- PREPARE      takes 274669 kb
- EXECUTE      takes 345730 kb
- PEAK         takes 345730 kb
===================================
Used Peak Memory : 177384 kb
- HWM after run  : 186024 kb
- HWM before init: 8640 kb
===================================

After this PR with #13654

$ ./Product/out/bin/onert_run -r 1 --shape_prepare [0,[4,299,299,3]] --shape_run [0,[3,299,299,3]] -m /home/nfs/model/inception_v3/inception_v3.tflite
Model Filename /home/nfs/model/inception_v3/inception_v3.tflite
===================================
MODEL_LOAD   takes 37.572 ms
PREPARE      takes 65.058 ms
EXECUTE      takes 107.753 ms
- MEAN     :  107.753 ms
- MAX      :  107.753 ms
- MIN      :  107.753 ms
- GEOMEAN  :  107.753 ms
===================================
RSS
- MODEL_LOAD   takes 105168 kb
- PREPARE      takes 146448 kb
- EXECUTE      takes 156376 kb
- PEAK         takes 156376 kb
===================================
HWM
- MODEL_LOAD   takes 110352 kb
- PREPARE      takes 146448 kb
- EXECUTE      takes 157236 kb
- PEAK         takes 157236 kb
===================================
PSS
- MODEL_LOAD   takes 194697 kb
- PREPARE      takes 278701 kb
- EXECUTE      takes 301738 kb
- PEAK         takes 301738 kb
===================================
Used Peak Memory : 153112 kb
- HWM after run  : 161944 kb
- HWM before init: 8832 kb
===================================

127.666 ms -> 107.753 ms
Peak memory: 177384 kb -> 153112 kb

[jyoungyun]

LGTM

[ragmani]

LGTM