gstlal-calibration:  lal_smoothkappas:  Changed algorithm to compute running median. Added property default-to-median. Also fixed memory bug in lal_smoothcalibfactors.

This may need to be improved in the future towards higher dimensions;
currently only uses information about m1m2. Only applies in stochastic
case.

- can now rely on scipy on LDG machines to supply this module

- fixes bug introduced with min_instruments that breaks XML I/O

- old function, no longer used, won't be used again

- property is no longer being used by anything except plotfar.py
- re-implement as function in plotfar.py

gstlal_compute_strain: Add median smoothing to all kappas and begin incorporation of coherence checking for kappa smoothing

- but do still yield (possibly empty) coinc params objects (if coinc_only is False)

- don't introduct bufdata variable
- speed up parsing of element name to get instrument

gstlal-calibration:  lal_demodulate:  Changed sign in exponential. Now it multiplies stream by exp(-i * omega * t).
                     lal_smoothkappas:  Changed default values for properties maximum-offset and kappa-ceiling.
                     calibration_parts.py:  Incorporated the property kappa-ceiling in function smooth_kappas. Added function smooth_kappas_real_test for testing various smoothing options. Used pow instead of complex_pow where appropriate.
                     gstlal_compute_strain: added option kappatst-real-ceiling and kappa-tst-imag-ceiling and used these in the smoothing of kappatst.

- reported by valgrind

- in service browser code

- better reflects the meaning of this paramter

- signal trapping has been removed from write_fileobj() and must be invoked
  directly by the calling code if desired (and we wanted it disabled in all
  cases here anyway)