Can intermediate rewards be used in reinforcement learning? Is it common practice in RL to have only one reward given at the end of the task?Or it is also possible to introduce subtasks/intermediate goals, so that feedback is not so delayed and more reward (functions) are necessary?
 A: 
Is it common practice in RL to have only one reward function awarded when a task is fulfilled in the end?

This isn't quite the correct definition of a reward function. An MDP has a single reward function, $R(s,a,s'): S \times A \times S \mapsto \mathbb{R}$, where $S, A$ are the sets of states, actions in the problem. You'll sometimes see versions with fewer arguments, say $R(s,a)$ or $R(s)$. 
$R$ returns rewards for every state transition. Many of them, or even all but one, can be zero. Or, other intermediate states can include positive or negative rewards. Both are possible, and dependent on the particular application. 
This the definition you'll find at the start of most reinforcement learning papers, e.g. this one on reward shaping, the related study of how one can alter the reward function without affecting the optimal policy.
A: I think the short version to your question is yes, it appears to be common practice to only reward an agent for full completion of a task, but be careful with your wording, as Sean pointed out in his answer that a reward function is defined for all possible combinations of states, actions, and future states.
To add to Sean's answer, consider these snippets taken from Richard Sutton and Andrew Barto's intro book on Reinforcement Learning:

The reward signal is your way of communicating to the [agent] what you want it to achieve, not how you want it achieved (author emphasis).
For example, a chess-playing agent should be rewarded only for actually winning, not for achieving subgoals such as taking its opponents pieces or gaining control of the center.

Although it does appear to be the recommended approach in their book, I'm sure you can find others who disagree.
A: If you're interested in subtasks, you want to look at options.  Aside from options, there is one reward function.
