Updated to read as_tibble by default (as per. #63). Still uses as_tibble = TRUE as an argument (rather than as.data.frame = TRUE)

Names are now all repaired using vctrs::vec_as_names, fixing #115 (something similar to the old scheme of repairs can be done using .name_repair = "minimal" although the letters as default column headings are now gone). This was unavoidable anyway though, so at least now they're different but consistent. Also this is what readxl does, so we should now match them. This adds vctrs as a dependency but as it's one of the dependencies of tibble, that's by the by.

Updated unit tests to check tibbles.

write_ods will naiively convert tibbles to dataframes before writing them, so that it is still possible to read and then write data without converting it outside the API.