[FEATURE] Support dynamic columns in PPL

Is your feature request related to a problem?
Certain PPL operators naturally require data-driven (dynamic) columns whose names and counts are not known at plan time. It is the blocker to support timechart and spath command

Example — timechart pivot

source=logs
| timechart span=1h count() by status limit=3

Problem: The set of status labels (e.g., ERROR, OK, TIMEOUT) is only known at runtime and can vary by dataset. Producing a wide, pivoted table requires dynamic columns, which are not available under a fixed per-operator schema.

Example — spath with object keys

source=idx
| spath input=doc path=event.tags

Problem: event.tags is a JSON object whose keys differ across documents. The extracted keys are unknown until read, so the output columns are inherently dynamic.

The issue names PPL support for dynamic columns and gives timechart and spath examples. Start by locating how PPL operators represent and validate output schemas, then trace the planning and execution paths for those operators. Done means the examples can produce runtime-defined columns without requiring a fixed per-operator schema.