Here are some benchmarks:

each_char:

    # /tmp/each_char.rb
    a = "a" * 1000000
    a.each_char do |x|
    end

Without this change:

    % time bin/mruby /tmp/each_char.rb
    bin/mruby /tmp/each_char.rb  1.07s user 0.02s system 99% cpu 1.088 total

With this change:

    % time bin/mruby /tmp/each_char.rb
    bin/mruby /tmp/each_char.rb  0.52s user 0.01s system 99% cpu 0.530 total

2 times faster with this change.

codepoints:

    # /tmp/codepoints.rb
    a = "a" * 1000000
    a.codepoints do |x|
    end

Without this change:

    % time bin/mruby /tmp/codepoints.rb
    bin/mruby /tmp/codepoints.rb  1.16s user 0.05s system 99% cpu 1.216 total

With this change:

    % time bin/mruby /tmp/codepoints.rb
    bin/mruby /tmp/codepoints.rb  0.56s user 0.02s system 99% cpu 0.589 total

This reverts commit 5cdcce8d.

define MRB_UTF8_STRING (in mrbconf.h) to enable UTF-8 support.

when non String object sent
And should be raise TypeError
when can not responsed to `to_str`