Fully Inlined call function, call-backward function.

[hikettei]

1. Optimized the overhead of call-forward/call-backward.
   All the generic functions are abandoned and now they're fully inlined.

Example:

That is, optimizing codes in cl-waffe has become very easy.
Various arguments were added to defnode accordingly.

:disassemble-forward 
:forward-declaim
:forward

:disassemble-backward 
:backward-declaim
:backward

(defnode ScalarSub ()
  :disassemble-forward t
  :forward-declaim (declaim (ftype (function (ScalarSub waffetensor waffetensor) waffetensor) :forward))
  :forward ((x y)
	    (let ((x (data x))
		  (y (data y))) ; todo: define with typevar
	      (declare (type single-float x y))
	      (const (- x y))))
  :backward ((dy) (list dy dy)))

TODO:

1. The defnode has been changed, but the standard implementation functions have not yet been optimised.
2. The main causes of overheads now are kernel.lisp And I need to optimise here.
3. Update Documentations.

⚠️ Some APIs are changed (e.g.: call now becomes macros), However, I don't have much time to update the documentations. So Documentation and API may differ.