Why cannot we take one slab of p-type semiconductor and physically join it to another slab of n-type semiconductor to get p-n junction?

By merely mechanically joining two slabs of n-type and p-type materials, it does not ensure good electrical continuity between the two slabs. Let me re-iterate that: mechanical continuity or connection does not always mean good electrical connection and continuity. For one, electronic properties of a material are a strong function of the material properties. Material properties including defects and other artefacts at the atomic level strongly affect the electronic properties. A monocrystalline piece of Silicon (Or, Germanium, or any other host material) has ideal properties to create a good P-N junction. It is simply impossible to create a monocrystalline material by joining two chunks of silicon like that. It is much harder to create the perfect electrically continuous P-N junction in this manner than one may realise.

Imagine aligning two materials such that there's perfect match in orientation at the atomic level to create a monocrystalline block of material! Even if we ignore the elephant in the room and assume that somehow we can manage to get the two crystal orientations aligned to be within tolerance limits somehow, there are other factors that hold us back from "joining" them. A quote from Wolfgang Pauli is super famous in the materials and devices community - "God made the bulk; the surface was invented by the devil". One can produce/manufacture near-perfect bulk material, and predict its behaviour so well with theory, but the material's surfaces, are full of dangling bonds, chemically active sites that are often terminated with undesirable functional groups, and host impurities such as adsorbants, and what not! While surfaces can be deterministically studied, surfaces are SO much harder to engineer to the specifications we will need to achieve a P-N junction. While the idea of simply putting two slabs together is appealing and sounds easy, the challenges involved in engineering the surface to achieve the kind of electrical continuity we want is simply impossible. Maybe we can get some sort of P-N junction behaviour by ignoring some of these technicalities, the devices so formed wouldn't be upto specifications we have been able to achieve using alternate methods. Besides, manufacturing would be a bigger nightmare! 

In the age of integrated circuits -- where we have BILLIONS of electronic devices and even more PN junctions in a square inch of monocrystalline silicon, manufacturing electrical components by processes such as "joining" n and p doped materials - it is simply not a scalable solution. We need scalable manufacturing techniques in those cases. But even for discreet elements which aren't a part of very large scale integrated circuits, we don't have to "join" two chunks of materials, because we have developed really good methods -- methods that are reproducable, controllable, and reliable -- to create P-N junctions in a single piece of silicon many decades ago. There is literally no reason for us to even attempt to join two disjoint pieces of n-type and p-type materials to create a P-N Junction. (For more information, you can read more about a technique called compensation doping, done through a process called ion implantation).