Atomic increment conflicts

[lann]

For databases where an increment uses optimistic concurrency under the covers, increments could fail due to conflicts in the same way as swap. With the current design I think explicitly signalling a conflict would require a new e.g. increment-error type, though if a bigger redesign is imminent this could be considered then.

[thomastaylor312]

We are already talking about improving CAS, so we should probably just add this error in at that point as well!