Add an option to stop before creating the vcf to allow filtering for noisy datasets that produce too many candidate SVs

[amcpherson]

We have some datasets that produce many candidate low read count SVs, and SVABA stalls when trying to deduplicate. If we stop SVABA during the dedup step we can then filter the .bps.txt.gz file and run svaba refilter to create a VCF in a reasonable time. This is somewhat manual and it would be preferable to run SVABA in a mode stops before creating the vcf allowing us to filter and run the last step of SVABA to dedup and create the VCF.