Unnecessary allocation of memories for channels in the JBM decoder

In the case of JBM operation, the output channels' buffers are allocated twice, once as st_ivas->p_output_f[n] and then as st_ivas->hTcBuffer->tc_buffer[]

A substantial memory saving can be achieved if these buffers are merged. (For the non-JBM path, the saving was in the order of tens to hundreds of kBytes as reported at #876 (comment 32711).)

Note: This is a follow-up to #858 (closed) where the same issue was solved for the non-JBM case.

Exemplary RAM consumption comparison between non-JBM and JBM decoder (commit 9da419d7 from April 8, 2025):

IVAS format		stack		heap
	non-JBM	JBM	increase [%]	non-JBM	JBM	increase [%]
EVS, 32 kbps	87,416	184,876	111	196,360	247,360	26
ISM, 32 kbps, 1 ISM	112,560	184,872	64	181,076	232,076	28
stereo, 32 kbps	132,584	184,872	39	264,596	346,552	31
SBA, 512 kbps, HOA3	166,408	184,736	11	884,646	1,083,926	23
OSBA 512 kbps, 3OA, 4 ISMs	166,408	184,736	11	1,456,946	1,780,050	22

All numbers correspond to EXT output configuration at 48 kHz.