Fix overflow on data duplication with elimination order

[wegank]

When computing Gröbner basis with respect to an elimination ordering, leadmons[0] is initialized as an array of size num_gb[0] * (bht->nv - st->nev).

msolve/src/msolve/lifting-gb.c

Lines 755 to 770 in 8f84271

	/* Leading monomials from Grobner basis */
	int32_t *bexp_lm = get_lm_from_bs(bs, bht);
	leadmons[0] = bexp_lm;
	int32_t len = bs->lml;
	num_gb[0] = compute_num_gb(bexp_lm, len, bht->nv, st->nev);
	int32_t *bexp_lm2 = NULL;
	if(st->nev){
	bexp_lm2 = calloc(num_gb[0]*(bht->nv - st->nev), sizeof(int32_t));
	for(int32_t i = 0; i < num_gb[0]; i++){
	for(int j = 0; j < bht->nv - st->nev; j++){
	bexp_lm2[i*(bht->nv - st->nev) + j] = bexp_lm[i*bht->nv + st->nev + j];
	}
	}
	leadmons[0] = bexp_lm2;
	}

However, duplicate_data_mthread_trace supposes that leadmons[0] is of size num_gb[0] * (st->nvars).

msolve/src/msolve/duplicate.c

Line 149 in 8f84271

const long len = num_gb[0] * (st->nvars);

This causes undefined behavior on Line 162 below. On one of my machines, the line causes segmentation fault on input_files/henrion5-qq.ms for j between the two values, with a probability of about 5%.

msolve/src/msolve/duplicate.c

Lines 158 to 164 in 8f84271

	/* leadmons_ori[0] has already been allocated*/
	for(int i = 1; i < nthreads; i++){
	leadmons_ori[i] = (int32_t *)calloc(len, sizeof(int32_t));
	for(long j = 0; j < len; j++){
	leadmons_ori[i][j] = leadmons_ori[0][j];
	}
	}

This PR fixes the issue on my machine.