Abstracting command return types

I'm finding my code is doing a lot of this:

```haskell
o1 <- shellStrict cmd1 empty
case o1 of
(ExitFailure, stderr) -> handle err
(ExitSuccess, stdout) -> do

o2 <- shellStrict cmd2 empty
case o2 of
(ExitFailure, stderr) -> handle err
(ExitSuccess, stdout) -> do

...
```

I'd like to abstract this and am wondering why Tuple is used for `procStrict`/`shellStrict` instead of the Either monad, as with `inprocWithErr`/`inshellWithErr`?

I'm not a shell programmer, so I very well may be missing the intended use case.

Here's a related discussion in the context of `shell`:
[Stack Overflow - Composing ExitCodes in Turtle.](https://stackoverflow.com/questions/33225573/composing-exitcodes-in-turtle-why-is-there-no-monad-monad-transformer-instance)

Start by comparing the return types of procStrict and shellStrict with inprocWithErr and inshellWithErr. Read the linked discussion about composing ExitCodes in Turtle to understand the existing trade-offs. Done means the project has an agreed direction for abstracting command-result composition, but the issue does not identify files or tests to change.

Written by the indexing model from the issue text.