* Refactored monitor/logger/alerter registration code
These objects are now more self-loading and just need to be
passed a dict of their config options.
* Changed to using packages for monitors, loggers and alerters.
* Fixed bug where initial failure time was always the most recent
failure time.
* Fixed a bug with the DiskSpace monitor which meant the free space
on non-Windows platforms was incorrectly calculated.
* Changed the intention of the DiskSpace monitior on non-Windows
platforms to measure the non-superuser free space.
* Fixed a bug which could mean the main loop ignored the interval
if an error occurred during tests.
* Better support for the dry_run setting on alerters
* BulkSMS sender now limits messages to 160 chars (otherwise BulkSMS
rejects them).
* Formatted times in alerts to not include microseconds
+ Monitors now know their own name
+ Alerters now have configurable time ranges for operation.
+ Alerters now have days of operation.
+ Alerters can hold alerts that occur out-of-hours and update you
if the monitor is still failed when they become in-hours.
+ Email alerts for failure now include downtime (which will be 0 for
an alert that fires immediately, but may be useful for alerters with
a limit).