My question is in the title: Why do physicists refer to irreducible representations (irreps) as "charges" or "charge sectors"?
For concrete examples, irreps are referred to as "charge sectors" just above Eq. (2.22) on page 7 of this paper about reference frames and "charges" in the 5th line from the end of the abstract (and then throughout) this paper on symmetric quantum circuits.
On one hand, the Hilbert space $H$ describing a physical system can be split up into a direct sum of subspaces that carry possibly different irreps. Physical states that live in these subspaces are confined to this subspace (by definition) when they are acted on by the representation of some relevant symmetry acting on $H$. On the other hand, I understand a charge as some quantity that is conserved in a system. Conservation and invariance are notions of the idea. So is calling an irrep a charge somehow getting at the fact that an observable is block diagonal with respect to the decomposition of $H$ according to the invariant subspaces?
Another point of confusion: Should I think of a charge as an operator (observable), an observable's eigenvalue, or a subspace of a Hilbert space? I think of a charge as a conserved observable quantity so it would make sense for a charge to be represented as an operator.