Character (multi-byte UTF-8) is destroyed when character spanning
`IO::BUF_SIZE` (4096 bytes) exist.

- Prepare file:

  ```ruby
  File.open("sample", "wb") { |f| f << "●" * 1370 }
  ```

- Before patched:

  ```ruby
  File.open("sample") { |f| a = []; while ch = f.getc; a << ch; end; p a }
  # => ["●", "●", ..., "●", "\xe2", "\x97", "\x8f", "●", "●", "●", "●"]

- After patched:

  ```ruby
  File.open("sample") { |f| a = []; while ch = f.getc; a << ch; end; p a }
  # => ["●", "●", ..., "●", "●", "●", "●", "●", "●"]

`IO#readline` and `IO#readchar` process in character units.

Support parsing a Regexp literal with 'o' option

Refactor `mrb_type` in `include/mruby/boxing_word.h`

`byteslice` creates 2 string objects. `_bufread` creates one, and
modifies the original buffer string, that is more efficient.

Use `%C` to `mrb_name_error` in `mrb_method_search`; ref 5890c7d5

The `bison` on MacOS does not support `%define api.pure`.
This reverts commit f7c9f1f7.

In the typical case, `mrb_funcall` invocation would be skipped.

Revert part of "Refine `tasks/toolchains/(gcc|clang).rake`" (07c6b7f0)

`-Wzero-length-array` is not a mandatory option, so the original is better.

Fix warnings for invalid C++ option with `enable_cxx_abi`; ref #3618, #4703

To stop warnings since C++ do not accept `-std=gnu99` option.

Refine `tasks/toolchains/(gcc|clang).rake`

- Make sure to specify `-std=gnu99` for C compiler flag.
- Make sure to specify `-Wzero-length-array` for C/C++ compiler flag (Clang).
- Extract similar codes.

1. `$/` and other Perl-ish global variables are not defined in ISO.
2. The current Ruby policy do not encourage those variables.
3. Those variables has global effect and can cause troubles.

Add paragraph mode to String#each_line in mrblib

The `Rakefile` was a leftover from the time it was an independent gem.