Inconsistent node memory usage metrics

**Describe the bug**
Inconsistency has been observed between memory statistics reported by `/cluster/node/{name}/usage` for an instance node and the memory usage reported by `/cluster/node/{name}/usage?runId={runId}` for a run on the same node. In some cases, the values differ significantly, with the run memory usage reported as several times higher than the node level memory.

Start by comparing the responses from /cluster/node/{name}/usage and /cluster/node/{name}/usage?runId={runId} for the same instance node, using a case where the values differ significantly. Trace how node-level and run-level memory usage are calculated, then verify that both endpoints report consistent memory values for the same node.

Written by the indexing model from the issue text.