Add `intercept` option to canned forecasters, or maybe some convolutional model

I was hoping to make/approximate a convolution-based forecaster & get its kernel with `arx_forecaster`.

E.g. something like
```r
fc <- cases_deaths_subset %>%
arx_forecaster("death_rate_7d_av", "case_rate_7d_av", args_list = arx_args_list(intercept = FALSE))
fc$epi_workflow %>% extract_fit_engine()
```

but there's no such setting, and it's not clear how to even achieve this. I think I got this working for quantile regression in an epipredict test somewhere, and there is maybe [this](https://stackoverflow.com/questions/69004818/how-to-fit-a-model-without-an-intercept-using-r-tidymodels-workflow) approach for `linear_reg` (and I guess also `quantile_reg`). But that seems like something that has to be done inside the canned forecaster itself.