ascii_replace: Replace invalid UTF-8 bytes.
In jkroes/FixEncoding: Search for and replace invalid UTF-8 bytes

Replace invalid bytes detected by check_column_encoding with valid ASCII or UTF-8 characters. Involves manual construction of a replacement vector (see rep_str).

1	ascii_replace(dset, enc_check_results, column_name, rep_str)

`dset`	A data.frame or data.table.
`enc_check_results`	A list returned by calling `check_column_encoding`.
`column_name`	The name of an element in the list returned by `check_column_encoding`, corresponding to a column header in `dset` where encoding issues were detected (or possibly false positives).
`rep_str`	A matrix. The number of rows should equal the length of `enc_check_results[[column_name]]`. The number of columns should equal the maximum number of invalid bytes sequences observed in any element of `enc_check_results[[column_name]]`. Strings within the character vector consist of either a single character to replace in the corresponding strings of `enc_check_results[[column_name]]`, or a random filler character or string. Because some elements of `enc_check_results[[column_name]]` may have more invalid byte sequences than others, and because the number of columns in the replacement matrix is equal to the maximum number of invalid sequences, some matrix rows may need dummy strings in order to be completely filled. This matrix must be manually constructed, as there is no method for guessing the proper ASCII or UTF-8 character to replace an invalid byte sequence. The function is only capable of replacing single columns at a time. To replace additional columns, the data.table returned by `ascii_replace` must be fed back into the function as the value of `dset`–likely with a different value for `rep_str`. While this may seem like an error-prone approach, remember that you can script your manual construction. A good idea is to use the same filler word throughout your matrix. Then call `grep` using that filler word as the value of the `pattern` argument and the result of calling `ascii_replace` as the value of argument \xodex. If the filler word is matched, it probably means that you missed a secondary, tertiary, etc. invalid byte sequence in one of your observations. (It's happened to me at least once!).