Prevent bin_file() -> is_utf8_well_formed() buffer overrun (#271)

When supplied with a file of random data >256 bytes, read()
would return a full buffer of data[256] to bin_file().
That function loops through the buffer, but calls is_utf8_well_formed()
with the full length of the buffer.  When looping on the last byte of the
buffer, is_uft8_well_formed() reads past the end.

This commit fixes bin_file() to only inspect the remaining bytes in the
buffer.

Found and tested with CheriBSD on an Arm Morello platform running with strong
memory safety

filename.c

@@ -477,7 +477,7 @@ bin_file(f)
  edata = &data[n];
  for (p = data; p < edata; )
  {
- if (utf_mode && !is_utf8_well_formed(p, edata-data))
+ if (utf_mode && !is_utf8_well_formed(p, edata-p))
  {
  bin_count++;
  utf_skip_to_lead(&p, edata);