Precision difference when updating numpy versions

[jjzieve]

Its suggested in the README to use numpy=1.11. But if its updated to a newer version (e.g. 1.15.2) regression tests will fail due to precision differences when outputting floats (e.g. GQ score).

Example:
`AssertionError: '1\t1158675\t1KG_1_1158675\tC\tCCTT\t.\tPASS\t.\tGT:GQ\t0/0:0.43702784\n' != '1\t1158675\t1KG_1_1158675\tC\tCCTT\t.\tPASS\t.\tGT:GQ\t0/0:0.437028\n'

• 1 1158675 1KG_1_1158675 C CCTT . PASS . GT:GQ 0/0:0.43702784
  ? - -

• 1 1158675 1KG_1_1158675 C CCTT . PASS . GT:GQ 0/0:0.437028`

Probable solution:

• Limit precision (either via rounding or output formatting) to a hardcoded (or user-specified value) and regenerate regression truth data.