BND alt alleles are incorrect - and an attached fix

[HillJamie]

I believe the code attached on this issue will fix most of #359 and #333

===============
Attached is a BAM file made to include the 6 breakends in the VCF spec section 5.4 "Specifying complex rearrangements with breakends".

The chromosomes have different names 0 <-> chr2, 1 <-> chr 13, 2 <-> chr 17 than in the VCF spec and are much shorter, but the last 3 digits of the breakpoint positions are approximately the same as in the spec. This makes it simple to compare a vcf produced by Sniffles2 on this data to the example output in the spec.

For convenience, the spec says that the following lines should be present:

2 321681 bnd.W G G]17:198982]
2 321682 bnd.V T ]13:123456]T
13 123456 bnd.U C C[2:321682[
13 123457 bnd.X A [17:198983[A
17 198982 bnd.Y A A]2:321681]
17 198983 bnd.Z C [13:123457[C

When I run Sniffles 2.0.7 with the command sniffles -i vcf_examples_sorted.bam --vcf old_output.vcf, the output vcf file contains the lines:
#CHROM POS ID REF ALT
0 6681 Sniffles2.BND.0S0 N ]2:1000]N # Equivalent to bnd.W
1 4456 Sniffles2.BND.0S1 N [0:6681[N # Equivalent to bnd.U
1 4457 Sniffles2.BND.1S1 N ]2:4982]N # Equivalent to bnd.X

I have made some simple changes that produce this instead:
#CHROM POS ID REF ALT
0 6681 Sniffles2.BND.0S0 N N]2:4982]
1 4456 Sniffles2.BND.0S1 N N[0:6681[
1 4457 Sniffles2.BND.1S1 N [2:4982[N

The coordinates in the alt allele of Sniffles2.BND.0S0 are now correct, and the format of the alt allele e.g. t[p[ is now correct in all cases.

I have only made changes to sv.py for single sample calling mode and have not investigated whether further changes are needed for other modes. The altered sv.py file is in the attached zip file. I have not tried to make the changes beautiful - for example, it would be nice to change variable names to make things clearer (maybe is_first -> is_donor, is_reverse -> is_mate_donor).

bndBug.zip

(As an aside: the other 3 breakends can also be called from the attached BAM file, but, because the supplementary alignments are hard-clipped, #409 must be resolved before these will be output)