Add support for reconciling disjoint hierarchies

This seems to be unstable for some reason (non-positive-definite covariance), perhaps the .id is being used in reconciliation in some way.

library(fpp3)
tourism_stretch <- tourism %>%
  stretch_tsibble(.init = 16)
fits <- tourism_stretch %>%
  model(ets = ETS(Trips)) %>%
  reconcile(ets_adj = min_trace(ets, method='wls'))
fc <- fits %>% forecast(h = 1)
fc %>% accuracy(tourism) %>%
  group_by(.model) %>%
  summarise(MASE = mean(MASE))

Start by running the R reproduction in the issue, from tourism_stretch through reconcile and forecast, to observe the non-positive-definite covariance failure. Review the reconciliation entry point and related tests, if present; done means disjoint hierarchies reconcile and the shown forecasting workflow completes without the instability.