Refactor CServer::MixEncodeTransmitData for performance

To me it seems as if CServer::MixEncodeTransmitData() has a lot of branches, is messy and likely worth optimizing. For example , we could likely also use a pattern like
```
vecfIntermProcBuf[k] += vecsData[i] * fGainL;
vecfIntermProcBuf[k + 1] += vecsData[i] * fGainR;
```

instead of

```
if ( ( i & 1 ) == 0 )
{
// if even : left channel
vecfIntermProcBuf[i] += vecsData[i] * fGainL;
}
else
{
// if odd : right channel
vecfIntermProcBuf[i] += vecsData[i] * fGainR;
}
```

https://github.com/jamulussoftware/jamulus/blob/3827902c594d54ccc5e3e8a73ff345877ca72a57/src/server.cpp#L914

As this seems to be the probably most important function in Jamulus, this warrants a close investigation