fix: handle NA values in separate_longer_position() (#1625)

[LeonidasZhak]

Problem

separate_longer_position() errors when input contains NA values, while separate_longer_delim() handles them correctly by preserving NA rows in the output.

library(tidyr)
df <- data.frame(x = c("ab,cd", NA))

separate_longer_delim(df, x, delim = ",")
#>      x
#> 1   ab
#> 2   cd
#> 3 <NA>

separate_longer_position(df, x, width = 2)
#> Error in seq.default(1, max_length, by = width): 'to' must be a finite number

Fix

Modified str_split_length() in R/separate-longer.R to handle NA values:

1. All-NA input: Returns a list of NA_character_ values
2. Mixed NA/non-NA input: Processes non-NA values normally, then reinserts NA values at their original positions
3. All-empty + NA input: Handles the edge case where all non-NA values are empty strings

Tests

Added two new tests in tests/testthat/test-separate-longer.R:

• separate_longer_position() handles NA values (#1625): Tests mixed NA/non-NA input
• separate_longer_position() handles all-NA input: Tests all-NA input

Validation

• All 1314 existing tests pass
• New tests pass
• Behavior is now consistent with separate_longer_delim()

Fixes #1625

[DavisVaughan]

Assuming this is the same as tidymodels/broom#1281 (comment)