Possible stale store is passed into update call for mutate

[jnutter]

If two mutation requests (2 create with sync ids that get updated to object ids on success in our case), only the last mutation saves.

[ryanashcraft]

See my comment: #37 (comment)

Maybe we should reconsider favoring the most recent version instead of reverting to the original. Although I'm hesitant to do this because it's not an insignificant behavioral change. Thoughts @ManThursday?

[trashstack]

If I'm understanding correctly:

• The existing code silently drops intermediate updates if the request fails.
• This change prevents the optimistic update being rolled back on a failure.

Neither case seems great but I think the latter is marginally better because it's easier to add code to manually rollback the optimistic update than it is to recover lost changes from other requests.

Perhaps we should merge this and then add an optional (or even required) "rollback" property to the query config so that there is an explicit way to handle the failure of an optimistic update. What do you think?

[ryanashcraft]

@jnutter Thanks for the PR and raising this issue. After consideration we decided we should definitely fix this issue, but we're going to do it within a larger set of changes that should also address #9. I plan to work on these changes soon, I'll be sure to mention you once I have a PR.