Implement frame generation dependent on emoji size
This makes frame the size of the picture that we are looking for.
This is better looking than having fixed-size frames.
In this commit image access is refactored into a function.
This implementation is so very slow because of the transparency issues,
but it is functional and without leaks.
I would like to refactor some of the functions as they git quite verbose
with the layered structures.
All the information need not be saved in 2 places.
Having to deal with the mlx image structure may not be as convenient,
but the math is quite simple so it should not be significant drawback.
Implement some basic behaviour of the window, such as closing.
So far the window just shows the input argument image.
I use minilibx-linux with a commit from myself, that solves
a memory issue.