`n_distinct` with `group_by` produces incorrect SQL for MSSQL

[carlganz]

library(dbplyr)

con_test <- simulate_mssql()

translate_sql(n_distinct(y), con = con_test, vars_group = "x")

Use of DISTINCT is not allowed with the OVER clause in MSSQL.

[hadley]

Minimal reprex:

library(dbplyr)

translate_sql(n_distinct(y), con = simulate_mssql(), vars_group = "x")
#> <SQL> COUNT(DISTINCT `y`) OVER (PARTITION BY `x`)

^{Created on 2023-11-02 with reprex v2.0.2}

Not clear what we can do here except throw a clearer error message.