Fix segfaults when assigning all-NA-vector to factor column

[shrektan]

Closes #4824

library(data.table)
dt = data.table(FACTOR = factor(rep("a", 100L))) 
set(dt, i = 1:99, j = "FACTOR", value = rep(NA, 99L)) 
dt
str(dt)

Before the PR, print(dt) leads to an error that Error in as.character.factor(x) : malformed factor. By inspecting the value of dt$FACTOR, you should find that some of the values are garbaged.

This PR fixes the issue.

[codecov]

Codecov Report

Merging #4829 (3d7f52c) into master (1f0d00e) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #4829   +/-   ##
=======================================
  Coverage   99.42%   99.42%           
=======================================
  Files          73       73           
  Lines       14465    14465           
=======================================
  Hits        14382    14382           
  Misses         83       83           

Impacted Files	Coverage Δ
src/assign.c	99.70% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1f0d00e...3d7f52c. Read the comment docs.

[mattdowle]

Great fix! Thanks for chasing it down.