I would love that too, but Python is a poor choice for mobile environments where every CPU cycle costs you battery life. Python code often takes up to 30x more CPU to accomplish the same work. I already get annoyed at how quickly my Moto X drains the battery when navigating or playing games; if that were 30x faster, the phone would be useless.
Now, you could argue that maybe Google should just make a more efficient Python interpreter. They did that for Android Java anyway, with Dalvik. But the design of the Python language makes it hard to optimize; witness the failure of Unladen Swallow, and all the incredibly talented engineer-hours that have gone into PyPy. A lot more happens, semantically, in each expression or function call than happens in Java or C.
CPython is not very fast, but that has nothing to do with Dalvik. When designing Dalvik from scratch they could have chosen any syntax, they didn't have to stick with Java. They could just as easily have chosen something Pythonic and ended up with something close to Groovy.
Performance was obviously not the main goal with Dalvik. If it was they would have gone with something like C++ or Go. The first versions of Dalvik was quite slow. Compact bytecode looks like the design goal which probably made sense since mobile bandwidth wasn't great at the time.
Right, they could've chosen any syntax, but they can't necessarily choose any semantics. Dalvik (and now ART) doesn't just look like Java, it behaves like Java, which means that if you're a Java programmer you can pick up Android basically immediately and have a good idea how it works. If they'd chosen a Python syntax but left out, say, keyword arguments and metaclasses and overloading of built-in operators, I doubt many Python developers would consider it the same language.
Now, you could argue that maybe Google should just make a more efficient Python interpreter. They did that for Android Java anyway, with Dalvik. But the design of the Python language makes it hard to optimize; witness the failure of Unladen Swallow, and all the incredibly talented engineer-hours that have gone into PyPy. A lot more happens, semantically, in each expression or function call than happens in Java or C.