add support for groups other than rank in plotAbundance

[Daenarys8]

This commit deprecates rank and adds support for other columns in rowData
ping: #100

[TuomasBorman]

Thanks, seems to work! However, I think we should think if this is needed as this makes the function more complex. The same can be achieved as follows which might be more transparent (I prefer that)

data(GlobalPatterns, package="mia")
tse <- GlobalPatterns

rowData(tse)$bacteria_group <- sample(c("gram-pos", "gram-neg"), nrow(tse), replace = TRUE)

# Works fine
plotAbundance(tse, group = "bacteria_group", as.relative = TRUE)

# The alternative apporach
tse_sub <- agglomerateByVariable(tse, f = "bacteria_group", by = "rows")
tse_sub <- transformAssay(tse_sub, method = "relabundance")
plotAbundance(tse_sub, assay.type = "relabundance")

Second thing is that, group argument name seems not to be used in anywhere else. It might be good to have same naming as in other functions because that might help users to understand what the arguments do.

Instead of changing the name, one option is to "hiddenly" support other grouping variables. The name could be rank, but it would still work with other variables also. As grouping by rank is the most common thing, that might make the function more simple to user

[TuomasBorman]

This naming is related to issue/PR that is open in mia. Wait until we have fixed that.

[TuomasBorman]

microbiome/mia#623

[TuomasBorman]

We decided to prefer that user does (relative) transformation outside the function. That is why as.relative was moved to "hidden" parameter. This means that also agglomeration should be done outside the function. --> I moved rank/group parameter as hidden parameter.