A coordinate on a dimension no data variable has breaks every query on the table #286

A Dataset with a coordinate on a dimension that no data variable has registers without error, but then every query on it fails.

import numpy as np, xarray as xr, xarray_sql as xql

ds = xr.Dataset(
    {"t": (("x",), np.arange(3.0))},
    coords={"x": [0, 1, 2], "y": [10, 20]},  # no data variable has `y`
)
ctx = xql.Context()
ctx.from_dataset("ds", ds.chunk())
ctx.sql("SELECT x, t FROM ds ORDER BY x").to_pandas()

On main, unchunked: registration succeeds, and the query raises

DataFusion error: FFI error: Execution error: Failed to call stream factory: ValueError: Dimensions {'y'} do not exist. Expected one or more of FrozenMappingWarningOnValuesAccess({'x': 3})

Chunked (ds.chunk(), as above), from_dataset itself raises TypeError: unsupported operand type(s) for -: 'NoneType' and 'NoneType' in _block_len.

The schema and row count come from every dimension of the Dataset (ds.dims), while the blocks come from the data variables' dimensions. y has no place in t's rows.

Expected: the table for t has columns x and t and three rows, as ds[["t"]] (which drops y) gives. A fix could register each dimension group as ds[var_names], as the mixed-dimension path already does.

#275 sends such a Dataset to the lazy reader rather than to an in-memory table, so it behaves as on main, with a TODO naming this issue.