diagnostics action timeouts

[michel-laterman]

The fleet-server issue elastic/fleet-server#3648 will add a new duration attribute to REQUEST_DIAGNOSTICS actions.

The client proto should support specifying this duration for actions (perhaps with google/protobuf/duration.proto) and support passing it to diagnostics hooks.
The (optional) CPU trace hook should use the duration specified, and user defined hooks should also be able to access the value.