[Regression] fread cannot read a file if it has unquoted fields containing \r

[st-pasha]

require(data.table)
fread("A,B,C\n0,,\n1,hello\rworld,2\n3,test,4\n")

produces

Error in fread("A,B,C\n0,,\n1,hello\rworld,2\n3,test,4\n") : 
  Line 3 has too few fields when detecting types. Use fill=TRUE to pad with NA. Expecting 3 fields but found 2: <<1,hello>>

Such scenario is not rare at all: even a mature CSV writer (such as pandas.to_csv or data.table::fwrite) may fail to look for \r characters within the field, and consequently fail to quote them under "auto" quoting rule.