`filter()` after `read_parquet_duckdb()` silently fails with factor columns

[CoryMcCartan]

If the parquet file has factor/dictionary columns, read_parquet_duckdb() runs without errors. However, attempting operations such as filtering on the factor column do not work. Interestingly, other verbs like count() do work.

Clearly related to #88, but I did not see documentation on the duckplyr website of the factor-variable limitation at all, and it seems like the issue is possibly an interaction of filter() and read_parquet_duckdb(), since count() works.

Reprex (apologies for the file size, didn't have time to trim further
county.parquet.zip

library(duckplyr)

d = read_parquet_duckdb("county.parquet")

count(d, contest)
#> # A duckplyr data frame: 2 variables
#>   contest        n
#>   <chr>      <int>
#> 1 house     194040
#> 2 president 181106
#> 3 senate    101726
nrow(filter(d, contest == "president"))
#> [1] 0

d2 = as_duckdb_tibble(collect(d))
count(d2, contest)
#> # A duckplyr data frame: 2 variables
#>   contest        n
#>   <chr>      <int>
#> 1 house     194040
#> 2 president 181106
#> 3 senate    101726
nrow(filter(d2, contest == "president"))
#> [1] 181106