feat: make DataFrame.__eq__ and DataFrame.__neq__ raise, point users towards nw.all

[MarcoGorelli]

For Polars, df != 0 is equivalent to df.select(pl.all() != 0)

In Narwhals, however, we just return a single scalar False:

In [26]: df = pl.DataFrame({'a': [1,2,3], 'b': [4,5,6], 'c': [7,8,9]})

In [27]: df != 0
Out[27]:
shape: (3, 3)
┌──────┬──────┬──────┐
│ a    ┆ b    ┆ c    │
│ ---  ┆ ---  ┆ ---  │
│ bool ┆ bool ┆ bool │
╞══════╪══════╪══════╡
│ true ┆ true ┆ true │
│ true ┆ true ┆ true │
│ true ┆ true ┆ true │
└──────┴──────┴──────┘

In [28]: import narwhals as nw

In [29]: nw.from_native(df) != 0
Out[29]: True

We could match Polars' behaviour here:

• for DataFrame, __eq__ should do select(nw.all()==0) and __neq__ should do select(nw.all()!=0)
• for LazyFrame, both __eq__ and __neq__ should raise

but I think this is a chance to raise an informative error message and push people towards nw.all. In the long run, that's the solution that'll get them the furthest, as expressions are composable