I've just signed up for StackOverflow RSS feeds on Synapse. The first feed was this post. I thought it was a recent one but just notice it is from 2 years ago, @Lurifaxel just posted an answer 4 days ago. I'll learn how to work around here ... eventually. :)
To the topic, database constraints like primary keys and foreign keys are resources to prevent bad client code to mess up data consistency. This is very popular (and almost mandatory) on OLTP databases.
In OLAP systems, where you load data either in batches or streams, you usually don't want this because it slows down the ingestion process. You usually rely on stage tables and CTAS techniques that will render the consistent version of your table.
OLTP transactions should not be considered in OLAP / BI system like Synapse. Which makes primary keys and foreign key constraints just irrelevant and not necessary.
In case you really need to create a primary key constraint in a table, Synapse pool can do that but it will not validate data already in the table. Only new data (insert/update) will be checked for duplication.