Standard way to retrieve data fails if no dependencies exist

The function `get_data_by_id`, which is also being used by `get_shaped_data_by_runid`, only looks for dependent variables. That means that a flat/dependency-less run can not have its data retrieved by this function. Is this really what we want?

Here is a small example:
```python
import numpy as np

from qcodes.dataset.data_export import get_data_by_id
from qcodes.dataset.measurements import Measurement
from qcodes import Parameter

x = Parameter('x')
y = Parameter('y')

meas = Measurement()
meas.register_parameter(x)
meas.register_parameter(y) # no data retrieved
# meas.register_parameter(y, setpoints=(x,)) # yes data retrieved

with meas.run() as datasaver:

for x_val, y_val in enumerate(np.random.randn(25)):

datasaver.add_result((x, x_val), (y, y_val))

data = get_data_by_id(datasaver.run_id)
print(data)
```

Start at the `get_data_by_id` function and reproduce the dependency-less measurement shown in the issue. Check how `get_shaped_data_by_runid` uses the same path; done means retrieving data from a flat run as well as from runs with dependent variables.

Written by the indexing model from the issue text.