Uniform mixed-source corpus selection with Serval - confirmation needed

[bhartmoore]

Now that mixed-source training is available in Scripture Forge, we need to confirm that the onboarding app implementation matches the Serval/SF implementation where teams will land.

Current Serval/SF implementation allows for only one corpus, and can be described as "mix the sources evenly where they both exist, and use the only source where only one exists, to cover the full training corpus." (see [(https://github.com/sillsdev/serval/issues/319#issuecomment-2176020032)])

We need to confirm that when using the onboarding app without a separate book selection for the (optional) second source, this is the way mixing happens.