Format sequence syntax:

  %[modifier]specifier

Modifiers:

  ----------+------------------------------------------------------------
  Modifier  | Meaning
  ----------+------------------------------------------------------------
      !     | Convert to string by corresponding `inspect` instead of
            | corresponding `to_s`.
  ----------+------------------------------------------------------------

Specifiers:

  ----------+----------------+--------------------------------------------
  Specifier | Argument Type  | Note
  ----------+----------------+--------------------------------------------
      c     | char           |
     d,i    | mrb_int        |
      f     | mrb_float      |
      l     | char*, mrb_int | Arguments are string and length.
      n     | mrb_sym        |
      s     | char*          | Argument is NUL terminated string.
      t     | mrb_value      | Convert to type (class) of object.
     v,S    | mrb_value      |
      C     | struct RClass* |
      T     | mrb_value      | Convert to real type (class) of object.
      Y     | mrb_value      | Same as `!v` if argument is `true`, `false`
            |                | or `nil`, otherwise same as `T`.
      %     | -              | Convert to percent sign itself (no argument
            |                | taken).
  ----------+----------------+--------------------------------------------

This change will increase the binary size, but replacing all format strings
with new specifiers/modifiers will decrease the size because it reduces
inline expansion of `mrb_obj_value()`, etc. at the caller.

Merge pull request #4607 from shuujii/fix-UTC-offset-representation-in-Time-to_s-on-some-environments

Fix UTC offset representation in `Time#to_s` on some environments; ref #4604

Use own implementation to calculate UTC offset on Visual Studio 2015 or
earlier or MinGW because `strftime("%z")` on these environments does not
conform C99.

`MRB_INT_MIN` is the only integer value that has no corresponding
positive integer value (i.e. `-MRB_INT_MIN` = `MRB_INT_MIN`).

Refine message to `skip` in nested `assert`

- I think "Info" is used only to `skip`, so change to "Skip".
- Changed the default value of `assert` and specify the argument explicitly
  at the caller of `assert` because it is unnatural "Assertion failed" is
  output even though the assertion doesn't fail.

== Example:

  def assert_foo(exp, act)
    assert do
      assert_equal exp[0], act[0]
      assert_equal exp[1], act[1]
    end
  end

  def assert_bar(exp, act)
    assert do
      skip
    end
  end

  def assert_baz(exp, act)
    assert do
      assert_equal exp, act
      assert_bar exp, act
    end
  end

  assert 'test#skip_in_nested_assert' do
    assert_baz 1, 1
  end

  === Before this patch:

    ?..
    Info: test#skip_in_nested_assert (core)
     - Assertion[1]
        Info: Assertion failed (core)
         - Assertion[1-2]
            Skip: Assertion failed (core)
      Total: 3
         OK: 2
         KO: 0
      Crash: 0
    Warning: 0
       Skip: 1

  === After this patch:

    ???
    Skip: test#skip_in_nested_assert (core)
     - Assertion[1]
        Skip: assert (core)
         - Assertion[1-2]
            Skip: assert (core)
      Total: 3
         OK: 0
         KO: 0
      Crash: 0
    Warning: 0
       Skip: 3

Fix "Warn if assertion is missing inside `assert`"; ref ff43b2b9

Fix Time#to_s encoding on Windows

strftime() on Windows returns locale encoding time zone for "%z" even
if MSDN says "%z" is "The offset from UTC in ISO 8601 format; no
characters if time zone is unknown" in MSDN:

https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/strftime-wcsftime-strftime-l-wcsftime-l?view=vs-2019

So we need to convert encoding of string from strftime().

Merge pull request #4407 from shuujii/add-assert_raise_with_message-and-assert_raise_with_message_pattern

Add `assert_raise_with_message` and `assert_raise_with_message_pattern`

Remove `Enumerator::Chain#initialize_copy`

Merge pull request #4603 from shuujii/define-plus-to-Enumerator-and-Enumerator-Chain-instead-of-Enumerable

Define `#+` to `Enumerator` and `Enumerator#Chain` instead of `Enumerable`

I think `Enumerator::Chain#initialize_copy` is unnecessary because CRuby
doesn't clone elements.

Drop dependency from `mruby-enumerator` to `mruby-enum-ext`

Fix the lack of precision for `Time`; ref d7435506

Drop dependency from `mruby-enum-chain` to `mruby-enum-ext`

  - `Time.local` and `Time.utc` are able to use with `MRB_INT16 + MRB_WITHOUT_FLOAT`.

  - `time_t` is converted directly from the Ruby object.

  - `time + sec` and` time - sec` are not affected by the precision of `mrb_float`.

    Similarly, calculations are possible with `MRB_INT16 + MRB_WITHOUT_FLOAT`.

Fix mruby-time with `MRB_WITHOUT_FLOAT`; ref d7435506

Also fix the misfeature introduced in 23783a44, that ignores newlines
between method chains.

Drop dependency from `mruby-array-ext` to `mruby-enum-ext`

Move `NilClass#to_h` to `mruby-object-ext` from `mruby-enum-ext`

Refine `Array#(permutation|combination) test`

- No guarantees about the order in which the permutations/combinations
  are yielded.

- Drop dependency on `Enumerator`.

When the size of Xorshift128 seed (`sizeof(uint32)*4`) is bigger than
ISTRUCT_DATA_SIZE, `Random` uses Xorshift96 instead.

Add encoding argument to `Integral#chr`