A bunch of enhancements for dealing with large tar archives

[zozlak]

This pull request contains following changes to the Tar class:

• Adds support for files bigger than 8GiB (see the note on handling out of basic format range values on the https://www.gnu.org/software/tar/manual/html_node/Extensions.html#Extensions). The helper functions were implemented using static public methods to make them testable without a need to create >8GiB files.
• Optimizes tar data writing by using a bigger write chunk size (512 bytes at once kills I/O performance on modern storage) and avoiding unnecessary pack() calls (only the very last 512 bytes block needs it).
• Adds the Tar::readCurrentEntry() method which allows reading tar entry content while iterating trough the archive with the generator returned by the Tar::yieldContents(). This allows efficient inspection of large tar archive content without need to extract it. In my particular case it is allows me to check backup consistency with code like

  $chunkSize = 1 << 20; // 1MB, whatever
  $tar = new Tar();
  $tar->open('some_archive.tgz');
  foreach($tar->yieldContents() as $i) {
    $hash = hash_init('some algo');
    while ($chunk = $tar->readCurrentEntry($chunkSize)) {
      hash_update($hash, $chunk);
    }
    hash_final($hash);
    if ($hash !== $refHashTakenFromSomewhereElse) {
      throw new Exception('inconsistency!');
    }
  }

  This code uses constant memory (depending on the $chunkSize only) and reads the archive only once no matter the tar archive size and format (compressed or not).

[splitbrain]

That's pretty cool. Thanks!