Adding and offset option for 0s and 1s when using the step_logit()

For logit transformation of values between 0 and 1, it may be a good idea to have an option to add/subtract a number (e.g., .0001) to 0s/1s to avoid -Inf/Inf.

This is something similar to the `offset` argument of step_log.

For instance,
```
step_logit(x, offset=0.0001)
```
first searches for 0s and 1s in variable `x` and adds/subtracts 0.0001, then applies the logit transformation.