1. The observer effect at the quantum level does feel like a CPU optimisation - the location of a particle is a probability field until measured. 1.1 In my experience these ‘optimisations’ recur at different levels - eg. economics, gravity, mass distribution, weather, even politics. Generalised models perform very well until you take local measurements - but they’re less scientifically provable.
2. Dark matter/energy feels like a hack rather than a waste of CPU resources - it’s a vague effect unmeasureable in the baryonic reality we inhabit. In the meantime it alters the structure of solar systems, galaxies and the observable universe itself and it’s not clear how.
3. Occum’s Razor actually works against your argument. If it’s possible for base reality to contain simulated universes then there is already an [almost?] infinite probability our universe is simulated, as base reality could potentially hold [almost?] infinite simulations. If entities within those simulations can also create their own simulations then the chance of our reality being base reality becomes vanishingly small.

Your statement “You can view our universe as a computer program, you can also view it as the universe.” concerns me. This is true until you start trying to analyse how the universe works, but then all kinds of weird things crop up.