functions with different numbers of arguments

Right now, if you do

```julia
m = QuickMDP(
...
reward = (s, a) -> s^2
...
)
```

```
reward(m, s, a, sp)
```
will not work.

Not sure what the best way to fix this is.

Start by locating the QuickMDP reward handling and the reward(m, s, a, sp) entry point. Determine the intended behavior when the configured reward function accepts fewer arguments, then add coverage demonstrating the supported call forms and confirming the desired result.

Written by the indexing model from the issue text.