Refactor mapping activation

[philippotto]

When a mapping is activated, mapping.ts will dispatch finishMappingInitializationAction once the texture is updated. However, that code is not executed in the tests. For that reason, handleSetMapping in mapping_saga.ts uses this code:

  if (process.env.IS_TESTING) {
      yield put(finishMappingInitializationAction(layerName));
    }

However, this IS_TESTING related code is not ideal and should be refactored in my opinion.

When SET_MAPPING is dispatched with the same mapping that is currently active (so, a no-op), mapping.ts will not dispatch a finishMappingInitializationAction, because the listenToStoreProperty listener will not trigger (as the state did not change). however, the reducer will set the mapping state to ACTIVATING.
To circumvent this, updateMappingWithMerge in proofread_saga.ts currently avoids dispatching the no-op action. however, this could be solved more elegantly. quoting from the code:

/* TODO #9064: in case setMappingAction is called with the same mapping
     * as already active, the reducer will set the state to ACTIVATING
     * but the listenToStoreProperty handler in mappings.ts will never be
     * triggered, because the callback is only called if the identity of the
     * watched property changes.
     * three possible solutions:
     *   a) avoid dispatching setMappingAction when the mapping did not change
     *      (this is the current solution here).
     *   b) Don't set the state to activating in the reducer if the mapping identity,
     *      did not change.
     *      (I feel like this makes the logic that controls the lifecycle of the mapping status
     *      more complicated?)
     *   c) Refactor the mappings.ts code so that it reacts to all setMapping actions.
     *      (for example, this could happen in a saga. this would also solve the problem
     *      that the mapping_saga currently dispatches finishMappingInitializationAction
     *      when IS_TESTING is true).
     *      <-- my favorite
     */