I'm a big fan of SUS - the System Usability Scale that Pablo mentions. However, it really only covers usability and to some extent learnability. It also relies heavily on the subjective opinion of users at some point after a usability session, where users are unlikely to get a complete coverage of a product, or to be able to fairly compare a product with other similar products.
A rounded review of a website will rely on both qualitative and quantitative metrics. For many aspects of a site, there are no quantifiable metrics, so a qualitative view has to suffice, for example in areas of credibility or quality of search.
The only real way to get an impression for a broad range of diverse criteria is for a manual deep dive across as much of the product as possible.
Still - that's likely to generate a subjective result on the part of the reviewer unless they have a wide range of experience with reviewing other websites. In other words, given that a metric or an impression or a judgement must sit somewhere on a spectrum between two extremes, then the reviewer must know what those extremes are in order to asses the site in question. This might come from experience, or by reviewing a bunch of other entries or looking at competitor websites in a given category.
That all means that it's next to impossible to have a consistent 'score' which you can give a website as an overall maetric. You can compare. You can build a relative impression. But you can't really say site A in a banking test scores 62, and site B in a gaming review scores 78, and contrast those metrics against each other sensibly. You might have a high expectation for the behaviour of forms on a banking website, but (maybe) not so much for a gaming website.
But what you can do is try to build some kind of fingerprint or profile for a type of website and compare like-with-like within the context of its peers.
Smahing Magazine has an article on heuristic website reviews in which they describe exactly this process, using radar diagrams to visualize relative scores across different criteria.
From that article, here are some examples:
A radar plot showing poor performance across all heuristic categories.
A radar plot showing a website that performs well in all areas but one.
To emphasize the differences in the heuristic measurements, overlaying one radar plot on the other:
Note that even with such artefacts as these, the outcome from reviews are closely tied to the reviewer(s) and their individual approach to testing or reviewing, the depth to which they explore, the breadth of interaction, their familiarity with similar products, and their own set of expectations that they bring.