Bug in concatenation

There seems to be a bug in the concatenation code surrounding charsets, which adds a character that prevents files from downloading correctly:

When concatenating files, if any of them has a BOM present (or one of the files is UTF-8 encoded or contains unicode), the BOM should be stripped, prepended to the output, and the encoding headers updated.

Each file is modified from here: https://github.com/Automattic/nginx-http-concat/blob/623aa7b4135783b847f56d1da1df988851518769/ngx-http-concat.php#L161

What’s needed is a check for a BOM, and if found, replace it with an empty string, and set a flag, and if that flag is found, prepend the BOM on the final output.

Start in ngx-http-concat.php around line 161, where each concatenated file is modified, and trace the charset and BOM handling. Verify the result by concatenating files with and without a BOM or Unicode content: the output should contain at most the required leading BOM, with encoding headers updated and no extra character that prevents downloads.

Written by the indexing model from the issue text.