fix wrong spanning deletion length with NON_REF of HaplotypeCaller

[TomofumiSaka]

Dear GATK team,

I found and fixed a bug in HaplotypeCaller that incorrectly lengthened the deletion by one nucleotide in the case of gVCF mode.
Like this vcf (this result is made from open data HG001/NA12878 sample).

chr21 10452597 . G <NON_REF> . . END=10452602 GT:DP:GQ:MIN_DP:PL 0/0:147:99:146:0,120,1800
chr21 10452603 . T C,<NON_REF> 255.64 . BaseQRankSum=-0.276;DP=156;ExcessHet=0.0000;MLEAC=1,0;MLEAF=0.500,0.00;MQRankSum=-7.712;RAW_MQandDP=487874,156;ReadPosRankSum=1.061 GT:AD:DP:GQ:PGT:PID:PL:PS:SB 0|1:136,16,0:152:99:0|1:10452603_T_C:263,0,5564,675,5628,6356:10452603:59,77,0,16
chr21 10452604 . G *,A,<NON_REF> 255.64 . BaseQRankSum=-1.417;DP=158;ExcessHet=0.0000;MLEAC=0,1,0;MLEAF=0.00,0.500,0.00;MQRankSum=-7.712;RAW_MQandDP=495074,158;ReadPosRankSum=1.037 GT:AD:DP:GQ:PGT:PID:PL:PS:SB 0|2:136,0,16,0:152:99:0|1:10452603_T_C:263,681,6518,0,5638,5564,681,6462,5634,6446:10452603:59,77,0,16
chr21 10452605 . C *,T,<NON_REF> 191.64 . BaseQRankSum=-0.601;DP=161;ExcessHet=0.0000;MLEAC=0,1,0;MLEAF=0.00,0.500,0.00;MQRankSum=1.656;RAW_MQandDP=505874,161;ReadPosRankSum=-0.404 GT:AD:DP:GQ:PL:SB 0/2:131,0,19,0:150:99:199,665,6662,0,4918,4665,652,6264,4868,6100:50,81,9,10
chr21 10452606 . C <NON_REF> . . END=10452611 GT:DP:GQ:MIN_DP:PL 0/0:154:99:150:0,120,1800
chr21 10452612 . G A,<NON_REF> 0 . BaseQRankSum=4.613;DP=166;ExcessHet=0.0000;MLEAC=0,0;MLEAF=0.00,0.00;MQRankSum=-7.647;RAW_MQandDP=530720,166;ReadPosRankSum=1.518 GT:AD:DP:GQ:PL:SB 0/0:146,13,0:159:12:0,12,4420,474,4719,5968:64,82,0,13

chr21 10452604 . G *,A,<NON_REF> and chr21 10452605 . C *,T,<NON_REF> are not covered any deletion clearly but treated as spanning deletion.

Command is following.
$java -jar $gatk HaplotypeCaller \ --reference Homo_sapiens_assembly38.fasta \ --input CNR0028194.gatk_best_practice.GRCh38.chr21_10451605-10453605.bam \ --output out.vcf \ --pcr-indel-model NONE \ -ERC GVCF \ -L chr21:10451605-10453605
Reference is GATK Bundle hg38.
Input bam is placed in https://pezycomputing-my.sharepoint.com/:f:/g/personal/sakai_pezy_co_jp/EuiUh7J-eCpOmA_Xkf3cOEwByf3lqKpm4N4FdYy7B5FCJA?e=dlZwDl

In recordDeletion method, NON_REF Allele is used for calculation of delesionSize as zero length allele.
I modified it to skip when NON_REF.