Screenshots are attached.
For our project we have requirements on total build time of each pipeline, from unit tests all the way to production. This feature will make the total build time very easy to read.
In my opinion manual triggers do not make too much sense because pipelines should be automatic and autonomous. If one stage or task in the pipeline fails for some reason, it's better to start a new pipeline build, (e.g. fixing test failures, committing new code which kicks off the pipeline), not resume from where it fails.
That being said, when the pipeline is restarted manually, new build time will be used for calculation.