Proposed title of this feature request

Decouple terminationGracePeriodSeconds from revision-timeout-seconds

What is the nature and description of the request?

The desire is to be able to explicitly set the value for `terminationGracePeriodSeconds`independently of the `revision-timeout-seconds` parameter.

Why does the customer need this? (List the business requirements here)

Currently, the value for `terminationGracePeriodSeconds` is set by the reconciler based on other parameters, such as `revision-timeout-seconds` and `max-revision-timeout-seconds`. In some scenarios it is desirable to be able to terminate the pods quicker for example to conserve resources while rolling out updates to AI workloads where the `revision-timeout-seconds` may be set to a very large value to account for slow responses from models.

List any affected packages or components.

knative serving