The default implementations of type predicate macros use `mrb_type`.
But `mrb_type` with `MRB_WORD_BOXING` isn't very efficient, so the new
implementations avoid `mrb_type`.

In lld linked programs, .rodata comes before .text, thus mrb_ro_data_p
will return false for strings in .rodata. Change the lower bound from
_etext to __ehdr_start to catch these cases. This works for ld.bfd, gold
and lld, and it does not have false positives even if .init_array does
not exist.

Remove the branch that uses _edata: strings in .data can be modified so
this is semantically incorrect. Delete the __APPLE__ branch (its
manpages say get_etext() and get_edata() are strongly discouraged).
.init_array has been adopted by most ELF platforms to supersede .ctors.

Neither _etext nor _edata is used, so rename MRB_USE_ETEXT_EDATA to
MRB_USE_EHDR_START.

Use `mrb_define_method` instead of `mrb_define_alias`

Singleton class definition do not introduce its own class variable scope
in CRuby/JRuby. So should mruby.

```
module Mod1
  class << Object.new

    C = 1
    @@cv = 1

    p Module.nesting,        # => [#<Class:#<Object:0x55cb16e60a50>>, Mod1]
      constants,             # => [:C]
      class_variables,       # => []
      Mod1.class_variables   # => [:@@cv]
  end
end
```

Fix typo in `Array#difference` document [ci skip]

Simplify arguments check in `String#index`

Also fix document about type of the first argument.

Fix `Enumerable#filter_map` without block; ref d380c7d2

Remove `mrb_get_args(mrb, "")`; ref 30f37872

We needed to preserve the original method name somewhere. We kept it in
the `env` structure pointed from aliased methods. #1457 and #1531 tried
to address this issue. But this patch is more memory efficient.

Limitation: this fix does not support `super` from methods defined by
`define_method`. This limitation may be addressed in the future, but
it's low priority.

Fix `Fixnum#(to_s|inspect)` argument specs

Before this patch:

  $ bin/mruby -e 'p 3.to_s(2)'
  trace (most recent call last):
    [0] -e:1
  -e:1: 'to_s': wrong number of arguments (1 for 0) (ArgumentError)

After this patch:

  $ bin/mruby -e 'p 3.to_s(2)'
  "11"

Shrink `mrb_get_args()`

Entrust "no block given" error to `mrb_get_args()`

Fix argument specs to `Enumerable`

Some error messages will be changed.

As a side effect, all specifiers now accept the `!` modifier.

In terms of specifiers, argv is never referenced after a method call as
shown in #3090.

Reduction of object code can be expected.

If you need to refer to argv after a method call in the same loop,
update argv after the method call.

`Enumerator#size` is not supported [ci skip]

`MRB_METHOD_TABLE_INLINE` was fragile. It requires `-falign-functions=n`.
On platform that uses higher bits of function pointers, you can use new
`MRB_METHOD_T_STRUCT` configuration macro.

This is partial `aspec` check that only checks `MRB_ARGS_NONE()`.

So you don't need `-falign-functions=2` anymore. Instead your platform
must not use higher bits of the pointer (true for most platforms).
If not, you have to use `struct mrb_method_t` version.

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 }
  # => ["●", "●", ..., "●", "●", "●", "●", "●", "●"]