StartRun handler throws an exception when either the job or attempt doesn't exist

[sentry-io]

This has only occurred once, but I think so effort should be put into making the query in StartRun.validate_job_reachable/1 more robust.

I suggest testing with different scenarios where the attempt and/or the job might not exist - or that the workflow for a given job is not associated with this attempt.

But Repo.one() may return nil for whatever reasons, so we should return an appropriately invalid changeset.

Sentry Issue: LIGHTNING-7J

Protocol.UndefinedError: protocol Enumerable not implemented for nil of type Atom. This protocol is implemented for the following type(s): DBConnection.PrepareStream, DBConnection.Stream, Date.Range, Ecto.Adapters.SQL.Stream, File.Stream, Function, GenEvent.Stream, HashDict, HashSet, IO.Stream, Jason.OrderedObject, List, Map, MapSet, Phoenix.LiveView.LiveStream, Postgrex.Stream, Range, Scrivener.Page, Stream, Timex.Interval
  File "lib/enum.ex", line 1, in Enumerable.impl_for!/1
  File "lib/enum.ex", line 166, in Enumerable.reduce/3
  File "lib/enum.ex", line 4387, in Enum.reduce/3
  File "lib/lightning/attempts/handlers.ex", line 47, in Lightning.Attempts.Handlers.StartRun.call/1
  File "lib/lightning_web/channels/attempt_channel.ex", line 150, in LightningWeb.AttemptChannel.handle_in/3
...
(3 additional frame(s) were not displayed)

(Protocol.UndefinedError protocol Enumerable not implemented for nil of type Atom. This protocol is implemented for the following type(s): DBConnection.PrepareStream, DBConnection.Stream, Date.Range, Ecto.Adapters.SQL.Stream, File.Stream, Function, GenEvent.Stream, HashDict, HashSet, IO.Stream, Jason.OrderedObject, List, Map, MapSet, Phoenix.LiveView.LiveStream, Postgrex.Stream, Range, Scrivener.Page, Stream, Timex.Interval)