Non-density ridglines (e.g., smoothed time-series)

Should we add an option to type_ridge() whereby, instead of enforcing a density calculation over x, we just take the smoothed y values across the raw values of x?

A canonical use case would be plotting time series. (This is actually the case that came up for me at work today, hence motivating the issue here.) SO ggrides example along these lines: https://stackoverflow.com/questions/63265979/ggridges-with-time-series-r

If implemented, we'd need to decide what smoothing function to use. While ksmooth seems a natural choice, I worry that (a) it produces too much variability and (b) it only supports a subset of the kernels that density does (so we'd need some internal logic to head off argument mismatches). Maybe we roll our own simple kernel smoother?

Start with the type_ridge() implementation and review how density handles x values and smoothing arguments. Compare the proposed time-series use case with the linked ggridges example, then assess ksmooth and the density kernel options before defining what the new option and its completed behavior should be.