The old v0 format:
This is used for older indexes, and indexes that did not pass through
zlib.

<headerline>
<512offset> <512length> <filename>
...

The headerline:

TARIX INDEX v<indexver> GENERATED BY <program and ver>

The old index version is 0, so a sample header line might look like:
TARIX INDEX v0 GENERATED BY tarix-0.0.1

The data lines:

Each data line is the 512-block offset in the tar archive in decimal, the
number of 512-blocks used by that file (not the same as the file length, but
how many blocks to seek forward to get to the next record, or how many
blocks to read out to get only that file), followed by the filename, and
then a newline.  512-blocks means that byte offset 512 in the tar archive
would be represented by a 512-offset of 1.

The new v1 format:
This is used for newer indexes that passed through zlib

<headerline>
<512offset> <actual offset> <512length> <filename>
...

The new format follows the same basic form as the old one.  The index
version is now 1.  The actual offset is the real offset in the output stream
where the zlib checkpoint was placed.  A zlib reset and seek to this address
will allow decompression to resume at the 512offset of that file.
