Avoid redundant LPC coefficient rescaling to Q12 to preserve precision

Basic info

• Float reference:
  • Encoder (float):
  • Decoder (float):
• Fixed point:
  • Encoder (fixed): 1c5efba7
  • Decoder (fixed):

Bug description

In the encoder, LPC filter coefficients are repeatedly rescaled to Q12 format, even though the sub-functions already handle different scaling internally. This redundant rescaling introduces unnecessary complexity and results in a loss of precision.

Ways to reproduce

Box folder: ...\Box_EXTERNAL_IVAS_BASOP_VERIFICATION\issues\issue-xxxx