Parametric size support isn't well-defined

For some computation graphs, the shape is dependent on value of the computation graph's parameters. One example of such a computation graph is:
```python
n = pt.make_size_param("n")
a = pt.make_placeholder("a", shape=(10,n), dtype="float64")
out = pt.squeeze(a)
```

Notice how the shape of `out` depends on the value of "n" passed to the program. However, we do not account for such shape specialization based on the values of the parameters.

Start with the parametric-shape example in the issue and inspect how Pytato represents size parameters and infers shapes. The issue names no files or tests; done would require an agreed definition of shape specialization based on parameter values, followed by implementation and coverage for this example.

Written by the indexing model from the issue text.