Implementation of new nodes creation

This issue involves several different issues that can/should be implemented at once:

• #319 (closed) proposing more flexibility on the priors
• #318 (closed) and #450 (closed) proposing Sensor constructors taking ParameterServer as argument
• #364 (closed) and #449 (closed) asking for checks on the user defined states
• #381 (closed) proposing a new way of initialize the states of SensorBase
• #452 (closed) proposing a new API for sensors,frames and landmarks to accept VectorComposite
• #376 (closed) proposing remove ExtrinsicsSize from the WOLF_SENSOR_CREATE macro and inferring it from problem size.

The proposal to address all these issues partially or completely is the following:

1. A new class Prior containing all information needed for applying a prior in a state block:

class Prior
{
  string   prior_mode;
  VectorXd state;
  VectorXd sigma;
  bool     fixed;
  bool     dynamic;
  VectorXd sigma_drift;
}

2. Sensors constructors will be of two types: taking the ParameterServer or taking Priors and Params. The second ones to be used in tests.

3. SensorBase ParameterServer constructor will be the responsible of extracting information to build the Prior, for that it will require the derived sensor keys (all the states that have to be asked to the param server). Here, some checks can be done (sizes and quaternion normalized).

4. In the YAML file, the user will have to define the priors in the following way:

sensor:
  [...]
  states:
    P:
      prior_mode: "factor"    # can be "initial_guess", "fix" or "factor" ("nothing" doesn't have sense..?)
      state: [0, 0]           # that would replace the current 'pose'
      sigma: [0.1, 0.1]       # Only required in case of "factor"
    O:
      prior_mode: "fix"
      state: [0]
    I:
      prior_mode: "factor"    # can be "nothing", "initial_guess", "fix" or "factor" (I don't know if "nothing" has sense...)
      state: [0, 0]           # Only required in case of "initial_guess", "fix" or "factor"
      sigma: [0.1, 0.1]       # Only required in case of "factor"
      dynamic: true           
      drift_sigma: [0.1, 0.1] # Only required in case of dynamic=true
    T:
      prior_mode: "nothing"
      dynamic: true