Improve binary handling for .a files

[Wilfred]

These are detected as application/x-archive, but currently it's being treated as UTF-16 with one error.

Might want to see if we can make UTF-16 detection even stricter as well (e.g. read more bytes, or try to read a whole number of characters)

[Wilfred]

Might also want to distinguish counts of replacement characters versus null byte characters in the logs.