feat(pipeline-run): add task-status and task-wait commands

[arseniy-pplx]

Summary

• add task-status to print the {task_name: status} map for a root execution
• add task-wait to poll until every child task is terminal
• resolve leaf state through container state and nested pipelines through graph-state reduction
• treat SUCCEEDED and SKIPPED as non-failure terminals; task failures print their summary to stderr and exit 2
• validate wait bounds and apply the deadline while resolving children so retrying state requests cannot cause an unbounded overrun
• change the shared status reducer used by the existing run-level status/wait: WAITING_FOR_UPSTREAM and UNINITIALIZED now count as active (a run is no longer reported terminal while a task waits upstream), mixed terminal aggregates reduce to the failure terminal ahead of SKIPPED/SUCCEEDED, and wait bounds now reject NaN/inf
• retry transient HTTP 5xx on the per-poll root details fetch during task-wait within the wait deadline, matching the child-state tolerance; unbounded waits give up after a few consecutive failures so a persistent outage still surfaces

Context

Run-level status and wait do not expose which task is still active or failed. These commands provide task-level observability for CI and automation without requiring callers to reconstruct the execution tree from raw API calls.

Testing

• uv run pytest tests/test_pipeline_runs_cli.py
• uvx ruff check on the touched source files
• uv lock --check
• git diff --check