enh: add `separator` argument to `read_csv` / `scan_csv`

I don't really want to align all `read_csv` arguments, but `separator` is probably a pretty fundamental one which we could align between backends
- duckdb: `delimiter`
- polars: `separator`
- pandas: `sep`
- pyspark: `sep`

(same for `has_header`)

EDIT: and `schema`?