* Add Hut & Bahcall (1983) style test for resonance.
* Add flag to fewbody input for continuing integrations after they are
  complete.
* Fix string handling to ensure that there is a newline character at the end of
  every string that needs it.  This is important for log entries in the output
  data stream.
* Test on Mac OSX.  Need -traditional-cpp?
* Play around with __inline__ thing?
* Must add tests in fb_is_stable_triple() and fb_is_stable_quad() to look for
  possible collisions that would cause the hierarchy to be unstable.  This is
  probably relevant only when the stars are huge, like red giants.
* Whenever the outer binary in a hierarchy can be done analytically for part of
  the orbit, it should be done, instead of doing it numerically.  Currently, an
  outer binary will only be done analytically when it can be done for the whole
  orbit, i.e. as long as the inner binaries will never be perturbed in the
  orbit.  One way to implement the new scheme is by collapse()'ing such an
  outer binary and giving it an expiration date so that it is turned back on
  when one of the inner binaries will be perturbed.
* Instead of using L and A and the Kepler equation to advance a Kepler orbit,
  possibly use Gauss's f(t,t_0) and g(t,t_0) functions and the position and
  velocity at t_0:
	r(t) = f(t,t_0)*r(t_0) + g(t,t_0)*v(t_0)
	v(t) = df/dt*r(t_0) + dg/dt*v(t_0)
  as described in Solar System Dynamics by Murray and Dermott, near pages
  30-40.  If this scheme is more accurate for very small or very large
  eccentricities it may be worthwhile.
* Perhaps incorporate Jamie Lombardi's "Make Me a Star" routine for stellar
  collisions?
* Do away with dynamic memory in ks_func() by putting all those matrices in 
  params?  This would probably speed things up.
* Remove gsl_matrix functions from fb_nonks_jac()?
