ParameterWithSetpoints with DelegateParameter as setpoints

The following setup is disallowed by qcodes:
```py
from qcodes.validators import Arrays
from qcodes.parameters import Parameter, ParameterWithSetpoints
from qcodes.dataset import Measurement

p=Parameter('p', vals=Arrays(shape=(10,)), get_cmd=lambda : np.arange(10))
ps1 = ParameterWithSetpoints('ps1', vals=p.vals, setpoints=(p,), get_cmd=lambda : np.linspace(0, 1, 10))
ps2 = ParameterWithSetpoints('ps2', vals=p.vals, setpoints=(ps1,), get_cmd=lambda : np.random.randn(10))

meas = Measurement()
meas.register_parameter(ps2)
# ValueError: Paramspec ps1 both depends on ['p'] and is depended upon by ['ps2']
# This was caused while adding these interdependencies {ParamSpecBase('ps1', 'array', 'ps1', ''): (ParamSpecBase('p', 'array', 'p', ''),)}
```
When instead `ps2` has a `DelegateParameter` as setpoints that points to `ps1`, the measurement works (but probably shouldn't according to the current logic):
```py
from qcodes.parameters import DelegateParameter
dp = DelegateParameter('dp', source=ps1)
ps2.setpoints = (dp,)

meas = Measurement()
meas.register_parameter(ps2) # p is just forgotten about here
```
Finally, I'm wondering if the dual dependency should not indeed be *allowed*. Take a `ParameterWithSetpoints` whose setpoints are maybe some voltages that are being ramped. We could want to attach time stamp information to the voltage ramp (maybe the measurement of the `ParameterWithSetpoints` is not continuous, but only during certain windows of the ramp), the natural way of doing which would be to make the voltage ramp a `ParameterWithSetpoints` itself, with setpoints given by the time stamps. I.e., `ps2 -> measured_data, ps1 -> voltage_ramp, p -> time_axis`. Another conceivable option would be to define the time stamps also as setpoints of the measured data parameter (`ps2`), but this is also illegal (two setpoint params imply a 2d setpoint grid rather than two independent setpoints).

The first question here would be if this is even possible with the current dataset structure.

Start at Measurement.register_parameter and the dataset dependency validation that produces the shown ValueError. Examine whether nested ParameterWithSetpoints dependencies can be represented without treating them as a two-dimensional setpoint grid, and whether the current dataset structure supports the proposed time-axis/ramp model. Done means a clear feasibility or design decision for this dependency model.

Written by the indexing model from the issue text.