Possible bug in explain_forecast without lag explanation

[martinju]

There seems to be a bug in the test "forecast_output_arima_numeric_no_lags":

shapr/tests/testthat/test-forecast-output.R

Lines 216 to 236 in a62005f

	test_that("forecast_output_arima_numeric_no_lags", {
	# TODO: Need to check out this output. It gives lots of warnings, which indicates something might be wrong.
	expect_snapshot_rds(
	explain_forecast(
	model = model_arima_temp,
	y = data[1:150, "Temp"],
	xreg = data[, "Wind"],
	train_idx = 2:148,
	explain_idx = 149:150,
	explain_y_lags = 0,
	explain_xreg_lags = 0,
	horizon = 3,
	approach = "independence",
	prediction_zero = p0_ar,
	group_lags = FALSE,
	n_batches = 1,
	timing = FALSE
	),
	"forecast_output_arima_numeric_no_lags"
	)
	})

A warning is thrown whenever predict_model is called. The warning is
Warning in matrix(): data length [2] is not a sub-multiple or multiple of the number of rows [3]
and comes from the following line:

shapr/R/model_arima.R

Line 36 in a62005f

xreg_pred <- matrix(newreg[i, ], horizon)

Althoug this is just a warning, I suspect it causes erroronous behavior.

@jonlachmann: Since you know this code the best, do you mind taking a look?