[Bug?] Custom voice alerts assume contiguous and ascending ordering in voice.ini about deviation

[eried]

That is why I just append new voices to the end and have a excel to create the file just in case. But it would be nice if the file supported non-contiguous

[TheRealMoeder]

I was never overly satisfied with the whole voice handling, but it grew out of limits due to unidirectional communication with dfplayer and ram limitations on some targets. If we use internal id's in the model files, those can't be edited manually anymore, which is also not a good solution.

[somewhatlurker]

The current implementation already uses internal ids in model files, they just happen to look like the real mp3 ids. Manual editing shouldn't get any harder after the change.

Although on the topic of causing new issues, I realise now that that naive fix could screw up model files for users who have non-standard edited voice files that unknowingly hit this bug.
That could be fixed by being able to read two types of value, such as HOLD0=200 for old loading behaviour and HOLD0=mp3-200 for new loading behaviour -- then old config can't be broken (I'm not sure what an appropriate prefix would be though; mp3- or ini- seem appropriate, but aren't very descriptive)

[somewhatlurker]

I went ahead and made a branch with this fixed: https://github.com/somewhatlurker/deviation/tree/model-voiceid-fix
Feel free to play around with it and see if you think it makes sense.

[TheRealMoeder]

Can you make a PR? It's always easier to have Travis checking things early on.

[somewhatlurker]

Sure