This could for example be achieved by using a message broker like RabbitMQ. You’d still receive all trigger events as they are sent from Pipedrive but instead of processing them directly add them to the queue in your Pipedrive workflow.
Then have a second workflow consuming this queue through the RabbitMQ Trigger node which allows you to configure the Parallel Message Processing Limit option ensuring only one execution takes place at once: