fix(oban): handle non-list stacktraces in Oban error reporter

[whatyouhide]

Problem

When an Oban job dies from a gen-call exit, Oban surfaces the {module, function, args} from the exit reason in the :stacktrace field of the [:oban, :job, :exception] telemetry metadata, and not a stacktrace list.

Sentry.Integrations.Oban.ErrorReporter.report/5 passes that value straight into Sentry's :stacktrace option, which is typed as a list. This results in an error (that we saw in production):

** (NimbleOptions.ValidationError) invalid value for :stacktrace option: expected list, got: {NimblePool, :checkout, [#PID<0.493.0>]}
    (sentry) lib/sentry/event.ex:186: Sentry.Event.create_event/1
    (sentry) lib/sentry.ex: Sentry.capture_message/2
    (sentry) lib/sentry/integrations/oban/error_reporter.ex:114: Sentry.Integrations.Oban.ErrorReporter.report/5

The handler raises, and :telemetry then permanently detaches it. The affected node stops reporting all Oban job errors to Sentry until it restarts, so the first such exit silently creates an observability blind spot.

The existing normalization at the top of report/5 only handled the empty-list ([]) case; any other non-list value fell straight through.

[whatyouhide]

Ah man @solnic just saw now that this overlaps #1064 (which I had not seen). Tests and comments here are a bit more thorough but up to you.

[solnic]

@whatyouhide no worries, great that we know what is causing it!