Convert to 16-bit signed integer
For infinite and NaNs values, Scilab and Matlab int16 return different values.
Matlab Scilab
A = `int16`_(inf)
A = 32767
A = `int16`_(-inf)
A = -32768
A = `int16`_(nan)
A = 0
A = `int16`_(%inf)
A = -32768
A = `int16`_(-%inf)
A = -32768
A = `int16`_(%nan)
A = -32768