Changes made to improve the function:

- Replaced some L_add_sat+Mpy_32_16_1 using Word16 by Madd_32_32 using Word32 (better accuracy)
- Computed often used shift operand only once outside the loop
- divided cldfb loop with inner IF-statements into 2 loops without inner IF-statements
- replaced smoothing constant 16-Bit values for 1/6 and 1/20 by 32-Bit values (better accuracy)
The output file is slightly different (9e-5), the WMOPS are reduced by 34 (max. 150 -> 116).