[non-BE]Codec Level Harmonization: Non-diegetic panning of mono to stereo is 3LKFS quieter
When rendering an EVS mono bitstream to stereo, the output loudness is 3LKFS quieter than the input. For example: ```sh # with mono pinknoise normalized to -26 LKFS ./IVAS_cod 128000 48 scripts/testv/pinknoise/MONO.wav mono.192 # the two commandlines below are equivalent ./IVAS_dec -non_diegetic_pan 0 48 mono.192 mono2stereo.wav ./IVAS_dec -evs STEREO 48 mono.192 mono2stereo.wav # Same result with testv ./IVAS_cod 128000 48 scripts/testv/stv48c.wav stv48c.192 ./IVAS_dec -evs STEREO 48 stv48c.192 tmp.wav ``` [Mono pinknoise test signal from `ci/loudness-measurements-pinknoise`.](https://forge.3gpp.org/rep/ivas-codec-pc/ivas-codec/-/blob/ci/loudness-measurements-pinknoise/scripts/testv/pinknoise/MONO.wav?ref_type=heads) | Input file | Input Loudness | Output Loudness | Difference | |------------|----------------|-----------------|------------| | MONO.wav (pinknoise) | \-25.95 | \-29.55 | \-3.6 | | stv48c.wav | \-19.20 | \-22.21 | \-3.01 | The factor of 0.5 for the center used by non-diegetic panning causes this loudness drop. If rendering to either the L or R channels (`left or l or 90->left, right or r or -90->right`) the loudness remains at the input level. To preserve loudness a tangent panning law or center gain of <span dir="">\\frac{1}{\\sqrt{2}}</span> would be expected.
task