[FLINK-28804][formats] Use proper stand-ins for missing metric groups

[zentol]

Resolve a few cases of undesired behavior when a schema attempts to access the metric group.
We shouldn't fail in that case but silently ignore it if we have to, as we do in other areas of Flink (because a lack of metrics should never fail a job).

[flinkbot]

CI report:

• 65e6735 Azure: SUCCESS

Bot commands

The @flinkbot bot supports the following commands:

• @flinkbot run azure re-run the last Azure build

[mas-chen]

Makes sense to me, I encountered this issue as well as in other Flink components.

It's a separate change--but what do you think about adding warning logs to UnregisteredMetricGroup to emit a log that says a metric may not be reported

[zentol]

It's a separate change--but what do you think about adding warning logs to UnregisteredMetricGroup to emit a log that says a metric may not be reported

I'm not a fan of warnings if there's nothing the user can do about it, especially when the chance of us fixing these instances are slim.