Eval supports model for float32 data type

Eval supports model for float32 data type

* Eval supports model for float32 data type

[Problem]
1. after input tensor info applied input configuration,
could copy data as much as configured.
2. it is necessary to check input configed pipeline with various model

[Solution]
added eval opt model_dtype to verify model in uint8 and float32

Signed-off-by: Juyeon Lee <juyeonne.lee@samsung.com>
* fix labels size

* convert uint8 img to float32

* change opt order

Change-Id: I7b35ca82f656a53bb0dff0607c6bd2fe546ef975