Rethink behaviour when minimum is unspecified in cell_rows() and cell_cols()

[jennybc]

cellranger::cell_rows(c(5, NA))
#> <cell_limits (5, -) x (-, -)>
cellranger::cell_rows(c(NA, 5))
#> <cell_limits (1, -) x (5, -)>

I'm not sure this lack of symmetry makes sense. Why does c(5, NA) mean '5 and up' (or whatever the caller wants it to mean) but c(NA, 5)means '1 through 5'?

Specifically for readxl, it seems like we should be discovering the lower limit from the data if the minimum is unspecified.

[jennybc]

I'm not sure what the answer is, but I will resolve this before next release.