When the Java compiler turns a class into bytecode, it must first look to a parent class. That is because the underlying implementation of classes is to point to the bytecode of the parent class - which holds the relevant methods and fields. Then it adds in pointers to the code of the child class functions - some of which are mandated by the 'implements' keyword.
Because the parent class must be compilable, it is easier if the compiler knows up front what that class is. Further, you can extend only one class but implement any number of interfaces. The compilation time climbs if the extends keyword can be intermingled amongst any number of implements instructions. Compilers want to fail as fast as possible to decrease dev time, so this choice is logical. Further, it helps you think clearly about the class for the same reason.
new
word order. – Veloz