Unlike Sentinel-2 L2A products, Sentinel-3 OLCI L1B products don’t come with a cloud cover layer.
A solution around that would be to come up with a combination of bands (such as a variation of NDSI) or a threshold method (e.g. based on B21, which is quite dark over land) that helps you detect clouds in order to form a binary mask and then only return values that don’t fall within the mask. These methods will be somewhat approximative since cloud screening is quite complex but might help remove some of the noise.
Alternatively, you could use a combination of OLCI and SLSTR, since SLSTR has thermal bands that can help with cloud screening, but then you would have to use the Statistical API.