Cedar-backup3

* Fix minor typos in user manual and source code documentation.
* Properly annotate code implemented based on Python 2.3 source.
* Add info within CREDITS about Python 2.3 and Docbook XSL licenses.
* Fix logging for boundaries values (can't print None[0], duh).

* Re-license manual under "GPL with clarifications" to satisfy DFSG.
* Rework our unmount solution again to try and fix observed problems.
- Sometimes, unmount seems to "work" but leaves things mounted.
- This might be because some file is not yet completely closed.
- We try to work around this by making repeated unmount attempts.
- This logic is now encapsulated in util.mount() and util.unmount().
- This solution should also be more portable to non-Linux systems.

* Attempt to deal more gracefully with corrupted media.
* Unmount media using -l ("lazy unmount") in consistency check.
* Be more verbose about media errors during consistency check.

* Make 'extend' package public by listing it in CedarBackup2/__init__.py.
* Reimplement digest generation to use incremental method (now ~3x faster).
* Tweak manifest to be a little more selective about what's distributed.

* Fix bug related to execution of commands with huge output.
* Create custom class util.Pipe, inheriting from popen2.Popen4.
* Re-implement util.executeCommand() in terms of util.Pipe.
* Change ownership of sysinfo files to backup user/group after write.

* In sysinfo extension, explicitly path to /sbin/fdisk command.
* Modify behavior and logging when optional sysinfo commands are not found.
* Add extra logging around boundaries and capacity calculations in writer.py.
* In executeCommand, log command using output logger as well as debug level.
* Docs now suggest --output in cron command line to aid problem diagnosis.
* Fix bug in capacity calculation, this time for media with a single session.
* Validate all capacity code against v1.0 code, making changes as needed.
* Re-evaluate all capacity-related regression tests against v1.0 code.
* Add new regression tests for capacity bugs which weren't already detected.