Simplify, improve and generalize (total rewrite?) time sequence management

[Patrick N. Raanes]

• At the moment, t (absolute time) is "too important" in the code,
  compared to k (the time index). For example,

  • dxdt have signature dxdt(x,t,dt) instead of dxdt(x,k,dt).
    But a common situation is that dxdt needs to look-up some
    property (e.g. parmeter value) from a pre-defined table.
    For that purpose, k is better suited that t because it is
    and integer, not a fload.
  • If models simply have the signature HMM.dyn(x,k)
    (i.e. ticker would only yield k and kObs)
    then variable dt is effectively supported.

• Change KObs to KObs-1 ?

• Highlight the fact that t=0 is special

  • There shoult not be any obs then
  • Some of the plotting functionality are maybe
    assuming that tt[0] == 0.

[Patrick N. Raanes]

Duplicate of #83