df.columns, df.index and s.index cannot be assigned to

Renaming every column in one line is one of the most common things anybody does to a pandas frame, and it does not work here.

```python
df.columns = ["x", "y"]
# AttributeError: property 'columns' of 'DataFrame' object has no setter
s.index = [9, 8]
# AttributeError: property 'index' of 'Series' object has no setter
```

Measured against the pinned pandas 3.0.3, all three of `DataFrame.columns`, `DataFrame.index` and `Series.index` accept an assignment, and a wrong length is a `ValueError` reading `Length mismatch: Expected axis has 2 elements, new values have 1 elements`.

This surfaced from the conformance suite in an indirect way that is worth writing down, because the label it arrived under made it look like a harness quibble and it is not. Seven L0 records fail with `'property', expected 'attribute'`. The suite classifies a member by asking whether the class attribute is an instance of the builtin `property`, and pandas' `columns` and `index` are `pandas._libs.properties.AxisProperty`, a Cython descriptor that is not one, so pandas is recorded as `attribute` and firepanda as `property`. That looks like the suite testing a CPython implementation detail.

The reason it is not is that the two kinds differ in the one way a user can feel. `AxisProperty` has a setter and a bare read only `property` does not, so the difference the suite is reporting is exactly this gap, seen through a vocabulary that does not have a word for it. The right fix is here rather than in the harness.

Seven names are affected. `DataFrame.columns`, `DataFrame.index` and `Series.index` are the ones that should take an assignment. `Index.dtype`, `Index.is_unique`, `Index.hasnans` and `Index.inferred_type` are `CachedProperty` in pandas and are read only there too, so those four are the harness vocabulary and not a gap, and they are worth separating out before anybody starts on this.

### What the setter has to do

Length is checked against the existing axis and a mismatch raises `ValueError` with pandas' wording above. Assigning to `DataFrame.columns` renames in place without touching any column data. Assigning to `DataFrame.index` or `Series.index` replaces the labels without reordering anything, which is the difference between this and `reindex` and is worth a test of its own, since getting it wrong quietly reorders a user's rows.

Related to [#254](https://github.com/tamnd/firepanda/issues/254), which is that `df.columns` answers with a Python list where pandas answers with an `Index`. These are separate but they touch the same accessor, and doing them together would mean one pass over the type rather than two.