Fix issue in read_bam implementation, it got the wrong contig name

[gamazeps]

pysam's rname actually returns the index of the contig instead of the contig name (which is poor naming on their end, but they deprecated that API).

We should use reference_name instead to get the contig name

Also added 2 integration tests for reading bam and sam files, None for cram because it would require a fasta and I don't have a dummy one to use.

Fixed a bug in the json serialization, where if the tag was a list it ended up not being serializable (probably an edge case, but the dummy samtools test file had one)

Also added filetype inference to be able to read sam and cram files

[nvictus]

LGTM. My only suggestion is to remove the OrderedDict but it's not urgent.

[gamazeps]

All done :)