feat(rt): add error handling workload_start

[if0ne]

In the current implementation HostWorkload::Error is unused. I think it should be used when workload_start fails. For now workload_start do not set HostWorkload::Error if workload initialization fails, and the workload status remains HostWorkload::Starting .

[if0ne]

CC @lxfontes

There are some inconsistency: workload_status returns error if the workload not found but workload_stop returns WorkloadStatus::Unspecified in the same case. Should we unify this behavior?

[lxfontes]

agree on workload_start.

for status & stop: they should indeed have the same behaviour.
stop is the odd one here as we encode too much information in the result so the caller ( operator ) can reflect status upwards:

• error: something wrong stopping the workload ( operator action: retry )
• unspecified: the input workload_id was not found ( operator action: ignore workload, continue )
• anything else: the actual workload state ( operator action: update )

we shouldn't be returning a sentinel ( unspecified ) to derive behaviour. I think we might be missing a WorkloadState here, namely WORKLOAD_STATE_NOT_FOUND

Both status & stop returning WORKLOAD_STATE_NOT_FOUND when the workload is not in the hashmap. This ensures callers can differentiate when a workload id is not there versus when the workload is in error state.