Split rendering in VoIP mode - support for mixed formats

Follow-up to #1119

For OSBA and OMASA bitstreams the audio output from the decoder is not correct with split rendering in VoIP mode. The corresponding test cases in tests/split_rendering/test_voip_be_splitrend_vs_binaural.py are currently commented out.

Further work is required to correctly handle the mixed-domain processing of objects/bed and the relative delays between them.

This is also an opportunity to look into optimisations regarding unnecessary conversions between TD and CLDFB in split rendering processing path in the decoder.