Picoscope refactor

[wirew0rm]

• switch to composition
  Do not use inheritance for the different Picoscope models but
  encapsulate the APIs for the different models without gnuradio
  dependencies.
  There are now three levels of abstraction:

  • PicoscopeXXX.hpp: wraps the picoscope C headers and macros in a C++ friendlier way. Has to implement PicoscopeImplementationLike
  • PicoscopeAPI: defines PicoscopeImplementationLike implements a wrapper for shared functionality which does not depend on gnuradio.
  • Picoscope.hpp: The actual GNURadio block.

• Add a qa_PicoscopeAPI unittest
  This allows to test the Picoscope abstraction without any dependencies
  on gr4 to allow fast testing and verifying API assumptions.

• remove atomic and blocking IO logic
  The previous logic was unsafe because it copied data outside of the
  callback in which copying is safe. Additionally by spawning its own
  threads in addition to the picoscope driver's own threads, it was
  heavy on CPU. The actual data transfer is handled by the Picoscope
  driver thread, and the picoscope's work function just polls the driver
  functions. In combination with the blocking scheduler, this should
  reduce CPU load (trading off against latency).
  Also does not reimplement internal block functions like workInternal.

• switch from std::vector<Port<...>> to std::array<Port<..>, N_Channels>

• Add simple picoscope cli application to test the low level API
  independent of gr4

• Support for digitial, AUX and EXTERNAL triggers
  Digital is intentionally simple and only allows a single digital
  input. AUX is only available on the PS600 and EXTERNAL is only
  relevant for the non MSO variant of the 3000A/5000A.

Open issues:

• testing should be extended massively and structured better
• tag matching still has some issues depending on settings for timeout and sample-rate.
• error handling
• investigate control of buffer sizes and polling rates to improve
  latencies. Ideally some of it can be done in the gr4 scheduler
  already.