Multi-byte I/O dose too match for text input in some locale.

For example, "," in numbers treat as decimal separator, not word separator.
This makes "1,234" as "1234", not "1" and "234".

Use "C.UTF-8" to avoid this problem.

* Use UTF-8 for multi-byte character I/O everywhare

This fix is required for "libstdc++".
Without this fix, "libstdc++" only works for 7bit ASCII.

It fixes https://github.com/microsoft/pict/issues/24 .

* Use current locale instead of fixed UTF-8 locale

It is good for non-UTF-8 (ANSI/OEM) encoding input.
Use environment variable "LANG" or "LC_CTYPE" to change input encoding.

(example)
$ LC_CTYPE="C.UTF-8" pict utf8-input.txt

* Add document for non-ASCII (UTF-8/ANSI) input encoding

Signed-off-by: Cleber Rosa <crosa@redhat.com>

Several classes and functions were not const-correct. Did the minimal changes
to fix compilation, although there are other const fixes that can be done.

- replaced for_each with range-based for loop
- added missing std:: scope
- for-loop temp now declared outside the loop to be in scope
- removed casts in calls to trim()