First of all, you can differentiate between static (essentially all players move simultaneously and only once) and dynamic (essentially non-static) games.
An extensive-form game is essentially a game tree. This form of presentation makes sense when looking at games where players move sequentially. However, you could also represent a simultaneous-move game in a game tree. It just doesn't add any insights over the normal form (matrix).
Game trees are a great tool to present dynamic games such as standard sequential-move games with perfect and complete information (player 1 moves first, player 2 observes the move and reacts, and so on), and they can also be used to present games of incomplete information or games of imperfect information.
Repeated games are dynamic games in which the same static stage game is repeated multiple times. Such games are dynamic but usually not represented in extensive-game form. If the stage game is not always exactly the same, but instead the game to be played in each period depends on a state, we speak of a stochastic game. How the state transitions from one state to the other can depend on the actions played before (and pure chance). Strictly speaking, A repeated game is a stochastic game with only a single state.
I have never heard the term "evolutionary game" without "theory" following. I am not sure if there is a definition of "evolutionary games". However, I see evolutionary game theory as a subfield of game theory that deals with certain forms of dynamic games and also different underlying assumptions. In particular, there is often no "common knowledge of rationality" and often players find their "optimal" strategy by trial and error without really knowing what they are doing.