Replaced the autopilot-specific inclusion of MAVLink to be more generic

Unittesting fix

Datalogger refactoring

I had to update some method calls in UASUnitTest.cc. I'd bet that some of the tests are wrong now because of it.

TODO: This should be altered to reuse stuff from qgroundcontrol.pro by including it instead of copy/paste.

Refactored the LogCompressor. It wasn't properly handling certain message logging files that mixed messages (Issue #70).
One note is that this logger should work correctly as I've tested it, but could have some edge cases. It is pickier about the file format that it will sparse, but I don't know if the error-checking stuff from the old code would have actually worked. This can also be easily re-added.

This code is also much faster than the old stuff. From what I could tell it scanned through the log file at least twice, I think three times. It also copied a lot of data into memory instead of reading, processing, and writing one line at a time, so memory use should be much lower. Some memory leaks from the old code were also refactored out, so lifetime memory use should be down.

Fixed a couple of uninitialized arrays within the MAVLinkDecoder. It would result in weird time values in some data logging instances.

Replaced the autopilot-specific inclusion of MAVLink to be more generic. Now if you have no actual QGC modifications to make and just want your messages to be available for viewing/logging you can just create the user_config.pri file and override MAVLINK_CONF.

Fixed a stack overflow bug when using the regression fitting command

Also changed calculateRegression(...) to return a bool, as that's all it needs.

Fixed a stack overflow bug when using the regression fitting command in logfile plot. Solution involved just allocating on the heap instead of the stack.

Linechart UI improvements

Canceling out of file selection in the logfile plotter no longer triggers a warning.

Added a check for a division by zero in AbstractCalibrator::logAverage()

Added a check for a division by zero in AbstractCalibrator::logAverage(). This was causing a crash when attempting a calibration when there's no RC data being received.

Fixed canceling the Logging function through the button on the main toolbar