Per https://github.com/dimitri/pgloader/issues/491 the 01208 code page for
IXF files is actually some kind of UTF-8. Let's see about that with this
change.

The IBM documentation saith it's "UTF-8 with IBM PUA" wherein PUA means
apparently Private Use Area.

http://www-01.ibm.com/software/globalization/ccsid/ccsid1208.html

We default to the file header code page in those cases, per
documentation, anyway.

For parsing the LOB Location Specifier, the blog post at the following
URL did help a lot, the IBM docs not mentionning what it should be
expected to look like...

  https://www.toadworld.com/platforms/ibmdb2/w/wiki/7545.exporting-lobs-and-identity-columns

The binary LOBs support is not expected to work fully yet for lack of
sample data. If you need this support, you should know how to help
getting it: send sample data my way!

From the specs about the IXFDRID field:

  The record identifier,  which identifies a particular  D record within
  the sequence of  several D records contributing to a  row of data. For
  the first D  record in a row of  data, this field has a  value of one;
  for the second  D record in a row  of data, this field has  a value of
  two, and  so on.  In each row  of data, all  the D  record identifiers
  called out in the C records must actually exist.

In this commit we make it so that the IXF reader is able to switch to
the next data record when needed. To get there, the internal APIs needed
to change, and that triggered in turn a review of the main API, which
has been improved.

Read: the main API just got broken, you need to change your code if you
use this lib.

Finding definitive docs about this Packed Decimal Format for a Fixed
Point numeric value is harder than it should, and got more hours of
attention that I'd like to admit.

Converting the specifications into code, one step at a time.