Fix memset bugs in bam2bcf_iaux.c and bam2bcf.c

In bam2bcf_iaux.c, the memset after the first pass was zeroing `aux`
(the nucleotide frequency counts) instead of `cns` (the consensus output
buffer). This caused the consensus-building loop to see all-zero counts
and produce 'N' for every position, discarding all insertion data.

In bam2bcf.c, remove a duplicate memset(call->SCR, ...) that was a
copy-paste error (two identical lines in a row).

Author: sirus20x6

bam2bcf.c

@@ -211,7 +211,6 @@ void bcf_callaux_clean(bcf_callaux_t *bca, bcf_call_t *call)
     if ( call->ADF ) memset(call->ADF,0,sizeof(int32_t)*(call->n+1)*B2B_MAX_ALLELES);
     if ( call->ADR ) memset(call->ADR,0,sizeof(int32_t)*(call->n+1)*B2B_MAX_ALLELES);
     if ( call->SCR ) memset(call->SCR,0,sizeof(*call->SCR)*(call->n+1));
-    if ( call->SCR ) memset(call->SCR,0,sizeof(*call->SCR)*(call->n+1));
     if ( bca->fmt_flag&B2B_FMT_NMBZ )
     {
         memset(call->ref_nm,0,sizeof(*call->ref_nm)*(call->n+1)*B2B_N_NM);

bam2bcf_iaux.c

@@ -228,7 +228,7 @@ static int iaux_init_ins_types(indel_aux_t *iaux)
         iaux->minscns = ncns;
     }
     else cns = iaux->inscns;
-    memset(aux,0,ncns*sizeof(*cns));
+    memset(cns,0,ncns*sizeof(*cns));
 
     // use the majority rule to construct the consensus
     for (t=0; t<iaux->ntypes; t++)