Tidy covariates does not remove redundant covariates if timeId is NA

When extracting temporal covariates either with temporalCovariateSettings or the newer temporalSequenceCovariateSettings covariates that are not temporal will get a timeId of NA. If the covariate is marked for deletion in tidyCovariates it´s supposed to be deleted by lines 107-108 in here:

 newCovariates <- newCovariates %>%
            anti_join(covariateData$deleteCovariateTimeIds, by = c("covariateId", "timeId"))

But since one of the variables to join on is NA nothing will happen. There is a field for joins "na_matches" to have NA values match each other but it seems to be only supported for dataframes, see here:

To match NA values, pass na_matches = "na" to the join verbs; this is only supported for data frames.

A small repro would be:

df <- Andromeda::andromeda(a=tibble(x=1:3, y=NA), b=tibble(x=1:1, y=NA))
df$a %>% anti_join(df$b, by=c("x", "y"))

The first line should be removed from df$a but isn't. I guess a possible workaround would be to only join on covariateId when timeId is NA, but I couldn't quickly figure out a good way to do that with an andromeda object.

Start in R/Normalization.R at lines 107-108 and trace tidyCovariates through the temporalCovariateSettings and temporalSequenceCovariateSettings paths. Run the Andromeda/tibble anti_join reproduction from the issue, then add a regression test covering covariateId matches where timeId is NA. Done means redundant non-temporal covariates are removed without breaking temporal joins.