[WIP] Implement Cancel restart/Kill restart feature

[slinkydeveloper]

Fix #895. Depends on #2335

The fundamental commit where semantics are implements is this one: 0edc593

Essentially the behavior is the following:

• Cancel and Kill behaviors are left untouched
• If the "restart" option is enabled when cancelling/killing, at the end of the invocation the state machine will simply reset the in flight invocation status, removing the pinned deployment id and resetting the journal (except the first entry, that is the input entry)

This commit exposes the feature both in Admin API and in CLI

[slinkydeveloper]

We talked offline about:

• Make sure if you cancel and retry, and then subsequently kill, then kill takes precedence and no retry takes place (to do in this PR) DONE
• Introduce something like "k8s resource event log" for invocation status with start/finished/retried events (in a followup) WILL DO IN IMMUTABLE JOURNAL

[slinkydeveloper]

At this point, this is outdated and we don't need it anymore. It's mostly superseded by #3360