Fix race condition calling workflow onComplete twice

[bentsherman]

Fix #6692

PR #4299 created a race condition where certain errors (e.g. a failed retry) can cause onComplete handlers (from a trace observer) to be called twice

This is because certain error conditions abort() the session directly from a separate thread instead of throwing an exception. The main thread still tears down the session via destroy(). Both methods call shutdown0() -> notfiyFlowComplete(), in this case concurrently.

The shutdown0() already has a flag to prevent adding new hooks after shutdown, so this PR simply expands the flag to also make the method idempotent. Doesn't matter for the shutdown hooks since they are drained, but it does matter for the trace observers.

[netlify]

✅ Deploy Preview for nextflow-docs canceled.

Name	Link
🔨 Latest commit	de2227e
🔍 Latest deploy log	https://app.netlify.com/projects/nextflow-docs/deploys/6a5a2427a3baee00089473b3

[bentsherman]

Ideally I'd prefer the runtime to always throw an exception instead of calling session.abort(), or interacting with the session at all. But the session-runtime coupling is a harder problem to tackle

I suspect the runtime is doing this when it's on a separate thread, to ensure the exception isn't lost. If we can establish a clean pattern for propagating exceptions from threads and apply it everywhere, that should greatly reduce the dependence on the session object

For now, I think this fix is sensible given the current constraints