Looking through errors in Honeycomb I noticed some of the instances
of builds failing due to an invalid Python version (in either their
`.python-version` or `runtime.txt` file) were due to leading/trailing
invisible Unicode characters such as zero width space / zero width
non-breaking space.

Previously the error message for those was hard to understand, since
the invisible nature of those characters mean they don't show up in
the message, eg:

```
 !     Error: Invalid Python version in runtime.txt.
 !
 !     The Python version specified in your runtime.txt file isn't
 !     in the correct format.
 !
 !     The following file contents were found, which aren't valid:
 !     python-3.11.9
...
```

The reason for this is that `sed` uses the current locale, and so
the `[:space:]` character class used by the sed call in
`utils::read_file_with_special_chars_substituted()` was previously
matching Unicode whitespace. (In contrast, the Bash extended regex
used when parsing the read contents does not.)

After this change, the unicode replacement character will now be
shown (like it already is for other Unicode non-whitespace characters
and ASCII control characters), eg:

```
...
 !     The following file contents were found, which aren't valid:
 !     ���python-3.11.9
...
```

See:
https://www.gnu.org/software/sed/manual/html_node/Locale-Considerations.html

GUS-W-20034162.