libthai th_brk is limited to ≤ 2Gi characters

The th_brk() function takes the length of the input as a size_t, but it stores breaks in an int array, thereby limiting possible hits to the first INT_MAX characters.

To fix, we need to, if we can, loop over the input in increments of INT_MAX and adjust the positions afterwards. We should also ping upstream about this problem.

th_brk() is deprecated, btw, but the replacement th_brk_find_breaks() has the same API issue.