Rendering with FastConv to BINAURAL_ROOM_REVERB uses BRIR convolution instead of HRTF

Bug description

While rendering with FastConv to binaural with synthetic reverb uses BRIR convolution instead of HRTF. This causes wrong output and significant computation complexity increase.

Ways to reproduce

Decoding of SBA at 256kbps to BINAURAL_ROOM_REVERB with dry room acoustics parameters (see rend_config_dry_long.cfg) should be very similar to output of BINAURAL output mode. It is very similar to the output of BINAURAL_ROOM_IR mode instead. Also complexity figures are very close to those of BINAURAL_ROOM_IR mode.

Possibly related to #750.