[Elastic Agent] Add feature flag to the system metricsets to control whether metrics failures mark agent as degraded

[cmacknz]

This is a follow up to #40542, where we made partial metrics collection keep the agent healthy to minimize customer disruption for non-fatal errors.

Once we are confident we have fixed the source of the most common partial metrics collection errors we should return to having agent report itself as degraded when metrics collection fails in our own tests as a quality gate. This allows new errors to be easily flagged to us before release.

The current known issues preventing us from doing this by default are:

• [Flaky Test]: TestLongRunningAgentForLeaks/TestHandleLeak – Metricbeat input status reporting makes Windows agent permanently degraded elastic-agent#5300
• [metricbeat/system module] should differentiate kernel and user-space threads #40537

We should add a feature flag to the system module or the system metrics that turns this functionality on or off, and turn it on in the Elastic Agent integration tests. Adding this as part of the system integration configuration allows it to be controlled through the configuration of the system/metrics input and system integration without needing changes to the Elastic Agent control protocol.

Acceptance Criteria:

• A module or metricset level configuration flag exists controlling whether non-fatal system metrics collection errors report as degraded or healthy.
• The configuration flag has the agent report as degraded on error in all Elastic Agent integration tests and all tests in https://github.com/elastic/integrations/tree/main/packages

[elasticmachine]

Pinging @elastic/elastic-agent-data-plane (Team:Elastic-Agent-Data-Plane)