In 2001 the GNU tar introduced support for large and negative numbers
(https://www.gnu.org/software/tar/manual/html_node/Extensions.html#Extensions)

This is required to handle files bigger than 8G.

So far there was no way to read the data from a file in an archive
without extracting it and extraction of a single file required rereading
of a whole archive. This commit changes the yieldContents() in a way it
does not skip to the next header entry before returning a current header
content. A position of the next header entry is remembered instead and
rewinded to only at the next next() call on the generator. This allows
to read the current entry content until the next() call. For that the
Tar::readCurrentEntry() method was added.

Tar::addData(): pad only the last block of data and write everything
else with just a single writebytes() call and without pack().

Tar::addFile(): move the read chunk size to a class constant.

A bunch of enhancements for dealing with large tar archives

bugfix modification time is incorrect