[meta] Retrain older models

[eu9ene]

A meta-issue to track retraining of the older models like Italian, Portuguese, French, German, Spanish etc.

We'll need to retrain them to incorporate the latest robustness fixes. Also open-source datasets have improved since we trained them.

Those are high-resource languages with massive amounts of available training data, so training them is a bit more expensive than a typical language from our last big training. We'll need to do the optimizations first and also fix as many quality issues as we can before retraining not to do this again in the near future.

### Tasks
- [ ] https://github.com/mozilla/firefox-translations-training/issues/215
- [ ] https://github.com/mozilla/firefox-translations-training/issues/719
- [ ] https://github.com/mozilla/firefox-translations-training/issues/453
- [ ] https://github.com/mozilla/firefox-translations-training/issues/216
- [ ] https://github.com/mozilla/firefox-translations-training/issues/238