Adding new accessor for Goertzel filter to GRC block. Thanks to Marcus Leech for the last three modifications.

Adding accessors to Goertzel filter GR block to call setparms from implementation file for (re)setting parameters.

* maint:
  Fixing up fft and constellation plot to better handle zooming.
  Fixing zooming capabilities. Time axis does not track any longer due to what looks like a matplotlib bug that resets the xlims after the zoom occurs.
  Fixing up other plotting tools for data read errors.
  Using .min and .max on scipy arrays instead of min() max(); seems to be more portable.
  Fix how the end of a file is handled.
  PFB resampler: fix it this way to avoid the signed/unsigned warning.
  PFB resampler: fixes bug where filter could be looking past the number of inputs.

Fixing zooming capabilities. Time axis does not track any longer due to what looks like a matplotlib bug that resets the xlims after the zoom occurs.

Under extreme circumstances, optfir might never produce an answer (atten>300), so this puts in a check on the ripple; if it gets too large, stop trying.

PFB channelizer can be specified without external taps. Uses optfir to generate an internal filter to cover the channel bandwidth; user can specify the attenuation of this filter if desired.

Updating the arb. resampler to use the optfir filter that provides better specificatiion of stopband atten.

Adding new example script for using the new PFB arbitrary resampler interface. One resampler takes user-generated taps and another resampler just takes the resampling rate. Both input and output signals are plotted.

Modifying blsk2 wrapper for PFB arbitrary resampler to allow the user to just specify the requested resampling rate without providing their own filter taps.

Taps are then generated inside hier_block2 to cover full bandwidth of input signal.

Optional attenuation parameter may be provided.

Changing API for gr_skiphead to use uint64_t for the offset instead of size_t (still unsigned). Fixes issue #304.

Also uses uint64_t's internally so everyone is always on the same type. This should not affect anyone's use of the block.

Fixing output types from tap_type to o_type in gr_single_pole_iir.h. Doesn't make a difference in the current uses of this class, but could in the future. Thanks to Achilleas Anastasopoulos for pointing this out.

Tweak LIBUSB m4 script to not check for 'usb_debug' symbol on Windows, because the symbol does not exist for that platform (only, it seems)