fix: fix mixture detection when Dominance contains NA

[vrahier]

Since the statistics refactor (#172), summary() with several versions can fail with:

Error: id variables not found in data: Dominance

stats_pairs() now binds all versions of a situation with bind_rows() before calling format_cropr() once. Before #172, each version was formatted on its own.

When only one version has a Dominance column on a sole crop (e.g. a previous model version giving Dominance = "0"), the bound data has Dominance = c(NA, "0"). detect_mixture() counted NA as a distinct value, so it saw two values and treated the situation as an intercrop. format_cropr() then added Dominance to the columns used to reshape the data. That column was dropped when keeping only the columns shared with the observations, which have no Dominance, so reshape2::melt() failed.

detect_mixture() now ignores NA values when counting distinct Dominance values