Annotate SomeException with Graphula seed

[pbrisbin]

Open Questions:

1. Is this feature worth changing the exception type thrown?
2. Is that change in exception type a major version bump?

You can see in my second commit the kind of thing our users might have to do. It's an unusual thing to be doing, but it will be broken by this "fix".

---

fix: annotate SomeException with Graphula seed

ad00327

The seed is useful to re-run in a deterministic way, which is why we
prepend it to any HUnitFailure exception messages. This commit adds
the same treatment to all other exceptions by catching SomeException
and using checkpoint to add it as an annotation.

Here is an example:

Before:

  uncaught exception: GenerationFailureMaxAttemptsToInsert
  GenerationFailureMaxAttemptsToInsert (Just "entity already exists by this key") School}

After:

  uncaught exception: AnnotatedException
  AnnotatedException {annotations = [Annotation @GraphulaSeed (GraphulaSeed (-8068470525722554834)),Annotation @CallStack [("checkpoint",SrcLoc {srcLocPackage = "graphula-2.0.0.0-ArtCvOVJMvo1gDWVYi6qKp", srcLocModule = "Graphula", srcLocFile = "src/Graphula.hs", srcLocStartLine = 272, srcLocStartCol = 8, srcLocEndLine = 272, srcLocEndCol = 18})]], exception = GenerationFailureMaxAttemptsToInsert (Just "entity already exists by this key") School}

Changing the thrown exception type is a risk, since any outer catches
our users (or other libraries) have in place may no longer catch it. But
using checkpoint and changing e to AnnotatedException e should be
tolerable, due to the fact that AnnotatedException.catch exists, which
catches both types at once.

That is why the output is relatively ugly, because we want to rely on
checkpoint and AnnotatedException, so this is what you get.

I also noticed that the existing code changes the HUnitFailure
constructor used (both Reason and ExpectedBotGot are turned into
Reason by way of formatFailureReason). Ideally, we would not do that
(and there's no real need), but I'm leaving it alone for now, erring on
the side of a smaller diff for this change.

Closes #44.

chore: update README to use annotated-exception's try

ad5f5b7

This is necessary because now the example throws an
AnnotatedException GenerationFailure and not a bare
GenerationFailure exception.

This is an example of the sort of change our users might have to make
because of the seed annotation feature.

[cdmren]

At this point I think it might make more sense to use the built-in exception context support from base (since base 4.20, so it's available in a stackage LTS) instead of annotated-exception.

Suggested change

	_ -> checkpoint (Annotation $ GraphulaSeed seed) $ throwIO ex
	_ ->
	throwIO $
	addExceptionContext (GraphulaSeed seed) $ -- guard this line in a CPP conditional
	ex

[pbrisbin]

Well that certainly sounds much better, and it doesn't change the exception type (I assume), so the downsides are gone. And with the CPP, we don't have to drop any support.

[pbrisbin]

I'm going to take a new swing in a new PR.