Having more size pools will allow us to allocate larger objects
through Variable Width Allocation.

I have attached some benchmark results below.

Discourse:
  On Discourse, we don't see much change in response times. We do see
  a small reduction in RSS.

  Branch RSS: 377.8 MB
  Master RSS: 396.3 MB

railsbench:
  On railsbench, we don't see a big change in RPS or p99 performance.
  We see a small increase in RSS.

  Branch RPS: 815.38
  Master RPS: 811.73

  Branch p99: 1.69 ms
  Master p99: 1.68 ms

  Branch RSS: 90.6 MB
  Master RSS: 89.4 MB

liquid:
  We don't see a significant change in liquid performance.

  Branch parse & render: 29.041 I/s
  Master parse & render: 29.211 I/s

This commit implements arrays on Variable Width Allocation. This allows
longer arrays to be embedded (i.e. contents directly follow the object
header) which improves performance through better cache locality.

Must not be a bad idea to improve documents. [ci skip]

Truly editorial fix for comments.  This works better with Emacs'
set-justification-full function. [ci skip]

RARRAY_AREF has been a macro for reasons.  We might not be able to
change that for public APIs, but why not relax the situation internally
to make it an inline function.

Assertions in header files slows down an interpreter, so they should be
turned off by default (simple `make`). To enable them, define a macro
`RUBY_DEBUG=1` (e.g. `make cppflags=-DRUBY_DEBUG` or use `#define` at
the very beggining of the file. Note that even if `NDEBUG=1` is defined,
`RUBY_DEBUG=1` enables all assertions.
[Feature #16837]
related: https://github.com/ruby/ruby/pull/3120

`assert()` lines in MRI *.c is not disabled even if `RUBY_DEBUG=0` and
it can be disabled with `NDEBUG=1`. So please consider to use
`RUBY_ASSERT()` if you want to disable them when `RUBY_DEBUG=0`.

To fix build failures.

Devs do not love "impl".