rcstagger Code

* Added a README file to describe this project, and an example configuration file (needs lots of work to cover all the appropriate features of the config file)

* Moved the 'kshots' branch to the trunk

* Starting to move the 'kshots' branch to the trunk

* Implemented and tested the child reaping methodology introduced into linux kernel 3.4.x - all daemons now report to rcstagger when they terminate, not the init process (pid=1). This makes rcstagger compatible with any linux-based rc system :}
* Implemented a parent process tracking system, so when parents terminate, we know which service they were attached to. This will either be a "notice" level log message or a "debug" level message, depending on the exit status (0 = debug, non-zero = notice)
* This project is almost ready for migration to sourceforge or similar. Given that I can't be the only one interested in such a solution, I believe that it would be beneficial to post this project there. Hopefully, someone can come up with some methodology to make this project not linux-specific (I definately wanted at least FreeBSD support for my own environment)

* Began work on piping output (stdout, stderr) to various log facilities (files, stdout, stderr, syslog) from any backgrounded service

* Fixed the process/daemon management - appears to be functional
* Added a dummy daemon to test with
* sockets are now closed when a daemon is forked out (in fact, all file descriptors get closed after forking now)
* Commands are again functional with the client (at least both listing commands and the stop command)
* Lots of refactoring to make things much cleaner and easier to follow

* Lots of refactoring - broke functions into different files by what they pertained to. Found some unused and nearly identical functions from this process...
* Implemented some typedefs for my structures so I can stop calling the "struct this" and "struct that"... now it's simply "this_t" and "that_t"

* Massive refactoring done in this commit
* Mostly refactoring the event handler - there shall only be one place in the daemon where messages shall be received, many where they may be sent. This will prevent deadlocks where two daemons are infinitely waiting for responses
* Added a message send/receive code chunk to the library section to standardize how our messages are sent and received
* Cleaned up a lot of local functions, declaring them "static" so they never reach the linker and slow down compile times (they're local - they shouldn't be linked outside their local files)
* Some of this will break the untested sibling communication code... but it's untested, and it would have deadlocked anyways (which prompted the design change)

* Lots of refactoring
* Added a "disabled" state to the service states, to prevent a given service from starting if it is disabled
* Added a switch for the prctl stuff, when it is (eventually) released
* Moved the service starting code into the server loop, so services will continually attempt to start on each event if they need to

* Updated the service_ctl code to properly track forks of services
* Added an "SERVICE_STATE_ERROR" state in case something goes wrong with a given service during a state transition