elm_macros: fix for lack of precision with double due to order of operation.

elm_macros: fix for lack of precision with double due to order of operation.

Summary:
The numerical formula below is calculated with the double type.
-> ELM_SCALE_SIZE(x) (int)(((double)(x) / elm_app_base_scale_get()) * elm_config_scale_get())

But if number of significant figure of the return value goes over 15, the epsilon error will happen.
Because of that, it is better to divide to do later.

Reviewers: woohyun, cedric

Reviewed By: cedric

Differential Revision: https://phab.enlightenment.org/D2900

Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>