Unify implementations of custom measurements and custom resets

[orpuente-MS]

This PR unifies the implementations of custom measurements and custom resets. It does two things:

• Removes the CallableKind::Measurement variant from HIR and FIR.
• Uses the Attr::Measurement to identify that a callable is a measurement.

[minestarks]

I don't know the motivation behind the change, so I can't judge it based on that, but the code changes look good to me.

[orpuente-MS]

I don't know the motivation behind the change, so I can't judge it based on that, but the code changes look good to me.

Just to leave some context documented here. When implementing custom measurements a couple of PRs ago, I encoded the fact that an operation was a measurement as the new (now removed) CallableKind::Measurement variant. It looked like a good idea at the moment because measurements are treated different in theory and literature.

In the next PR when adding reset operations, an edge case came afloat, the MResetZ operation, which is a measurement and a reset at the same time. So, taking @swernli suggestion, instead of adding more variants to the CallableKind enum, I just piped down the attributes of a callable (@Measurement, @Reset, etc.) and used that information to determine if an operation was a reset, which happened to be simpler and more scalable when adding new attributes in the future.

In this PR I just undid the addition of the CallableKind::Measurement variant and used the infrastructure already added for resets, the piped-down attributes, to determine if an operation is a measurement.