Bug in PairParameter -- can't be fit!

[paulray]

WAVES are a new kind of parameter with a pair of values. I just tried to use it with fitting and it fails like this:

Traceback (most recent call last):
  File "ex_pn.py", line 52, in <module>
    f.fit_toas()
  File "/Users/paulr/src/PINT/src/pint/fitter.py", line 177, in fit_toas
    fitpv = self.get_fitparams_num()
  File "/Users/paulr/src/PINT/src/pint/fitter.py", line 97, in get_fitparams_num
    for k in self.model.params
  File "/Users/paulr/src/PINT/src/pint/fitter.py", line 98, in <genexpr>
    if not getattr(self.model, k).frozen
  File "/Users/paulr/src/PINT/src/pint/models/parameter.py", line 1165, in value
    return self.param_comp.value
  File "/Users/paulr/src/PINT/src/pint/models/parameter.py", line 1729, in value
    return self.get_value(self._quantity)
  File "/Users/paulr/src/PINT/src/pint/models/parameter.py", line 619, in get_value_float
    return quan.to(self.units).value
AttributeError: 'list' object has no attribute 'to'

This is demonstrated by examples/ex_pn.py in PR #587

I was trying to fit a model with WAVES to some data with pulse numbers as part of my fixes for pulse numbering, which didn't seem to work before.

[paulray]

Note that in addition to this, PINT counts degrees of freedom by the number of parameters with frozen=False, and this will be wrong for PairParameters. We need to rework WAVES to generate two parameters, I think. But then things like as_parfile() will need to know that the two parameters are coupled and write them out on one line.

[scottransom]

We talked about the PairParameters issue a bit before on the telecons. I personally think they should go away. As for parfile lines, It would probably be really simple to have the parfile readers/writers combine those if necessary/requested, but we could write them one per line as a default.

[abhisrkckl]

pairParameters were introduced for the Wave component. It will be superseded by the WaveX model (#1609).
We are not planning to fix this.