Predictors of treatment allocation guesses in a randomized controlled trial testing double-blind injectable hydromorphone and diacetylmorphine for severe opioid use disorder