Fix number of scalines read from GAC POD files

[oembury]

Fixes #136 by changing the gac_pod offset, plus some changes made for investigating:

• Added an option to disable call to correct_scan_line_numbers
• Updates pod version of correct_scan_line_numbers to return results from base class
• Adds code to discard the final scanline if we have:
  • two logical records per physical record (i.e. gac data)
  • an odd number of expected scanlines and actual length is exactly n+1 lines

Note - the final check is currently only discarding the expected padding at the end of a file. It will still read any unexpected padding and attempt to parse as scanlines. For production use it would probably be better to only read self.head["number_of_scans"] scanlines.

[sfinkens]

@mraspaud Does this have any effect on LAC reading?