Faster preselection on zero-width and 'negative'-width

[PitWD]

It would be my first contribution to a project which is not my own. So, no clue how to do.

I've forked the project: https://github.com/PitWD/termux_wcwidth The changes are made inside the wcwidth() function. And I added two functions to test_wcwidth.c for a simple time measurement.

Please check the changes and if you are fine with them, I would love to get a little introduction how to prepare them for a regular commit.

1st edit: ~10% faster. Binary size is a little smaller, RAM usage the same. 2nd edit: Even RAM usage is a little smaller...