linear regression: non dimension coords along dim end up in params

[mathause]

data_1D.year ends up in lr.params - but they should not

import mesmer
import mesmer.testing

data_1D = mesmer.testing.trend_data_1D()
data_1D = data_1D.assign_coords(year=("time", data_1D.time.values + 1850))

lr = mesmer.stats.linear_regression.LinearRegression()
lr.fit({"not_time": data_1D["time"]}, data_1D, dim="time")
lr.params

<xarray.Dataset>
Dimensions:        (time: 30)
Coordinates:
    year           (time) int64 1850 1851 1852 1853 1854 ... 1876 1877 1878 1879
Dimensions without coordinates: time
Data variables:
    intercept      float64 -0.07344
    not_time       float64 0.9967
    fit_intercept  bool True