v0.4-beta (Experimental)

Changes in this release:

• Add alpha_max + beta_min magnitude approximation. Should be faster than previous method, but both are available via -fastmag flag.
• Add benchmark for magnitude functions. Run with go test -bench=Mag. These are useful since the new method is slower on my Core i7 3770k and faster on my Atom D525. Your mileage may vary.
• Update travis-ci configuration since libFFTW is no longer required.
• Add license preamble to all source files.
• Refactored LogMessage for clarity.
• Add -cpuprofile flag to write pprof data during execution.

Future:

• Dumping IQ samples to file needs to be non-blocking. Will need to do some profiling to determine if sync.Pool is useful here since a copy of the IQ buffer will need to be made for each pass when a packet is decoded properly.