bug(dfir_rs): a subgraph may run multiple times in one tick due to race-ish behavior between the graph and external events

[MingweiSamuel]

Found while testing cross_join implementations, assert!(#context.is_first_run_this_tick()) fails. A subgraph should only run twice in a tick if there is a loop causing it to be re-scheduled (a fix point loop).

[MingweiSamuel]

This seems to actually be an issue with the dfir_rs scheduling mechanism, not hydro/paxos

[MingweiSamuel]

So what happens is a subgraph A lazily rescheduled itself due to replay (of fold_keyed in this case), then the next time that stratum is run, the subgraph runs itself, then the rest of the stratum runs and triggers a second run of A