asciimat() should be moved into [m2sci] module. It wrongly manages empty strings and hypermatrices of ascii codes

Reported by Samuel GOUGEON (@sgougeon)

BUG DESCRIPTION:
----------------
asciimat() is a pure internal Matlab conversion function. There is no asciimat() function neither
in Matlab nor in Octave.
Almost all its calls in Scilab are from functions of the M2SCI converter,
mainly from convert2double() that is called by almost all sci_functions.
The very few remaining calls outside [m2sci] are artificial and can be replaced by trivial
ascii() equivalent calls.
This is why asciimat() should be moved into SCI/modules/m2sci/macros/kernel, and its page to 
SCI/modules/m2sci/help.
A true distinct Scilab asciiMat() could be implemented for Scilab string objects of distinct lengths
and without horizontal concatenation, if it is required.

asciimat() is aimed to comput an equivalence of the Matlab/Octave +/- operations between char strings
-- not texts -- and numbers, as in
>> ['abc';'def']+0
ans =
    97    98    99
   100   101   102

But to this respect, the present asciimat() implementation is bugged:

1) The empty char '' is wrongly managed:
--> asciimat(['abc';'';'def'])  // yields a transposed result
 ans  =
   97.    98. 
   99.    100.
   101.   102.

while with Matlab as with Octave:

>> ['abc';'';'def']+0
ans =
    97    98    99
   100   101   102

2) When converting an hypermatrix of ascii codes to strings, the dim#2 of the result is wrongly squeezed:
--> t = cat(3,['abc' 'd';'e' 'fgh'],['ij' 'kl';'mno' 'p'])
 t  = 
(:,:,1)

  "abc"  "d"  
  "e"    "fgh"
(:,:,2)
  "ij"   "kl"
  "mno"  "p" 

--> c = asciimat(t)  // OK
 ans  = 
(:,:,1)
   97.    98.    99.    100.
   101.   102.   103.   104.
(:,:,2)
   105.   106.   107.   108.
   109.   110.   111.   112.

--> asciimat(c)     // KO
 ans  =
  "abcd"  "ijkl"
  "efgh"  "mnop"

// Expected:
 ans  = 
(:,:,1)
  "abcd"
  "efgh"
(:,:,2)
  "ijkl"
  "mnop"

as with Matlab and Octave:

>> t = cat(3,['abc' 'd';'e' 'fgh'],['ij' 'kl';'mno' 'p']);
>> c = t+0
c =
ans(:,:,1) =
    97    98    99   100
   101   102   103   104

ans(:,:,2) =
   105   106   107   108
   109   110   111   112

>> char(c)
ans =

ans(:,:,1) =
abcd
efgh

ans(:,:,2) =
ijkl
mnop



ERROR LOG:
----------
None. Wrong results


HOW TO REPRODUCE THE BUG:
-------------------------
// 1) The empty char '' is wrongly managed:
r = asciimat(['abc';'';'def'])  // yields a transposed result
and(r==[97 98 99 ; 100 101 102])

// 2) When converting an hypermatrix of ascii codes to strings, the dim#2 of the result is wrongly squeezed:
t = cat(3,['abc' 'd';'e' 'fgh'],['ij' 'kl';'mno' 'p']);
c = asciimat(t)
r = asciimat(c)
and(size(r) == [2 1 2])


OTHER INFORMATION:
------------------
See also the bug 16438 about supporting UTF-8 characters and input strings of distinct lengths.