core: be more selective about what we log

[eckter]

The logs generated by core can be flooded by a few things, which makes log less readable (and increase costs in prod)

The main thing to focus on is warnings when loading an infra. It can generate a large amount of log entries (2k2) and they're mostly ignored in normal workflows.

We may also look into warnings that are raised during simulations and stdcm requests (mostly margin related), though it's not necessarily excessive.