fix(k2proc): fix memset size calculation

[tachibana-shin]

bug: memset() only clears a fraction of the allocated pixel_count_array, leaving most of it uninitialized from malloc.

// Before (bug):
memset(pixel_count_array, 0, sizeof(int)*(region->c2+2) + (region->r2+2));
//  = 4*(c2+2) + (r2+2) bytes — only clears ~1/(c2+2) of the array

// After (fix):
memset(pixel_count_array, 0, sizeof(int)*(region->c2+2) * (region->r2+2));
//  = 4*(c2+2)*(r2+2) bytes — clears the entire array
Evidence:
1. Allocation uses * correctly (k2proc.c:1904):
willus_mem_alloc((double **)&pixel_count_array,
    sizeof(int) * (region->c2+2) * (region->r2+2), funcname);
2. Loop accesses the full 2D range (k2proc.c:1917-1933):
rows_per_column = region->r2 + 2;
for (i = 0; i <= region->c2+1; i++)
    cp = &pixel_count_array[i * rows_per_column];  // offset = i * (r2+2)
    for (j = 1; j < jmax; j++)  // jmax <= r2+2
3. memset must zero the same (c2+2)*(r2+2) ints that the loop writes to.

<!-- Reviewable:start -->
- - -
This change is [<img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/koreader/libk2pdfopt/70)
<!-- Reviewable:end -->

[Frenzie]

If you have the time, please also try to report those upstream: https://www.willus.com/email.shtml