Fix a couple small VCF auto-indexing bugs.

1. sam_idx_save wasn't validating the file is BGZF.  It's invalid
   usage to try calling this function on uncompressed data, but we
   should double check.

   Note this is triggered by a bcftools bug where -o
   foo.vcf.gz##idx##foo.vcf.gz.csi writes VCF rather than VCF.gz as
   the "filename" doesn't end in .gz.

2. Add the hts_idx_amend_last calls to vcf_write as we did previously
   for SAM/BAM.

   This isn't technically a requirement, as all it's doing is changing
   virtual offsets to an alternate form that gives the same file
   offset (see comments above hts_idx_amend_last), but doing so means
   the auto-build indices match those produced by a standalone index
   command.

   This fix isn't complete as it hasn't been worked on for BCF yet.
   However it comes under the "nicety" category and isn't really
   fixing a bug so we can try to figure out how to tidy up BCF later
   (plus VCF.gz is basically the universal format).

sam.c

@@ -1067,7 +1067,7 @@ int sam_idx_save(htsFile *fp) {
             errno = -ret;
             return -1;
         }
-        if (bgzf_flush(fp->fp.bgzf) < 0)
+        if (!fp->is_bgzf || bgzf_flush(fp->fp.bgzf) < 0)
             return -1;
         hts_idx_amend_last(fp->idx, bgzf_tell(fp->fp.bgzf));
 

vcf.c

@@ -3574,6 +3574,8 @@ int vcf_write(htsFile *fp, const bcf_hdr_t *h, bcf1_t *v)
     if ( fp->format.compression!=no_compression ) {
         if (bgzf_flush_try(fp->fp.bgzf, fp->line.l) < 0)
             return -1;
+        if (fp->idx)
+            hts_idx_amend_last(fp->idx, bgzf_tell(fp->fp.bgzf));
         ret = bgzf_write(fp->fp.bgzf, fp->line.s, fp->line.l);
     } else {
         ret = hwrite(fp->fp.hfile, fp->line.s, fp->line.l);