Improve serialization of IndexedDataTable

[abeham]

IndexedDataTables are stored extremely inefficiently in the serialized xml and take a huge amount of space.

A simple graph of 3 points (6 double values) requires more than 700 characters

<COMPOSITE name="values" typeId="30" id="79">
  <COMPOSITE name="list" typeId="31" id="80">
    <COMPOSITE typeId="32" id="81">
      <METAINFO>
        <PRIMITIVE typeId="13" name="Item1">4.7700000000000001e-05</PRIMITIVE>
        <PRIMITIVE typeId="13" name="Item2">943176159</PRIMITIVE>
      </METAINFO>
    </COMPOSITE>
    <COMPOSITE typeId="32" id="82">
      <METAINFO>
        <PRIMITIVE typeId="13" name="Item1">7.7999999999999999e-05</PRIMITIVE>
        <PRIMITIVE typeId="13" name="Item2">943175954</PRIMITIVE>
      </METAINFO>
    </COMPOSITE>
    <COMPOSITE typeId="32" id="83">
      <METAINFO>
        <PRIMITIVE typeId="13" name="Item1">9.3899999999999992e-05</PRIMITIVE>
        <PRIMITIVE typeId="13" name="Item2">943175905</PRIMITIVE>
      </METAINFO>
    </COMPOSITE>
    ...
  </COMPOSITE>
</COMPOSITE>

[abeham]

r15917:15918: implemented more efficient storage format

[abeham]

After these changes I tested with a large experiment containing more than 50,000 runs. Each containing an indexed data table. Compressed storage space was reduced from 87,605 kb to 49,710 kb. Uncompressed storage space was reduced from 1538 Mb to 831 Mb.

Deserialization is "quite" fast, but serialization is very slow.

[mkommend]

Reviewed r15918 (r15917 is only the branch creation).

Changes look good and are ready for trunk integration and subsequent release.

[abeham]

r16177: merged to trunk

[abeham]

r16178: removed branch

[abeham]

r16282: merged to stable