[Jobs] Show pool health in the wait spinner for pool-targeted launches

[lloyd-brown]

Summary

When sky jobs launch --pool <pool> can't dispatch immediately (e.g. no idle replicas), the job sits in PENDING showing only the generic "Waiting for task to start" spinner. The user has no signal whether the pool is fundamentally broken (zero READY workers) or just busy with someone else's work — they have to run sky jobs pool status in another terminal to find out.

This PR adds a one-line pool-health summary to that spinner, refreshed each tick, so the situation is visible inline:

⠋ Waiting for task to start (status: PENDING). It may take a few minutes.
  Pool 'foo': 1/3 ready (2 provisioning)
  View controller logs: sky jobs logs --controller 42

• A small helper _get_pool_health_summary in sky/jobs/utils.py queries serve_state.get_replica_infos and produces strings like Pool 'foo': 5/11 ready (3 provisioning, 2 starting, 1 failed).
• All FAILED_* replica statuses (FAILED, FAILED_CLEANUP, FAILED_INITIAL_DELAY, FAILED_PROBING, FAILED_PROVISION, UNKNOWN) collapse to a single failed count via ReplicaStatus.failed_statuses() so the line stays readable for large/chaotic pools.
• A new {pool_str} field is added to _JOB_WAITING_STATUS_MESSAGE. When pool is unset (non-pool launches), it stays empty, so non-pool behavior is unchanged.
• Any error querying replicas is swallowed; the spinner falls back to the original message.

Test plan

Renderings across pool scales (verified via unittest.mock.patch.object(serve_state, 'get_replica_infos', ...)):

Scenario	Rendered summary
Empty / nonexistent pool	Pool 'demo-pool': 0 workers
Healthy small pool, all ready	Pool 'demo-pool': 3/3 ready
Small pool warming up	Pool 'demo-pool': 1/3 ready (2 provisioning)
Small pool, broken (mixed FAILED_*)	Pool 'demo-pool': 0/3 ready (3 failed)
Mid pool (mixed)	Pool 'demo-pool': 5/11 ready (3 provisioning, 2 starting, 1 failed)
Large pool (50)	Pool 'demo-pool': 30/50 ready (10 provisioning, 5 starting, 3 not_ready, 2 failed)
Chaotic 200-pool	Pool 'demo-pool': 80/200 ready (50 provisioning, 30 starting, 10 not_ready, 10 pending, 5 shutting_down, 15 failed)

End-to-end manual test:

• sky jobs launch --pool <healthy-pool> -- echo hi — confirm spinner shows Pool '<name>': N/N ready while waiting, transitions to streaming logs after dispatch.
• sky jobs launch --pool <pool-with-no-ready-workers> -- echo hi — confirm spinner shows 0/N ready (...) instead of leaving the user in the dark.
• sky jobs launch -- echo hi (no --pool) — confirm spinner is byte-identical to pre-change output (no pool line).
• format.sh clean (10.00/10 pylint, mypy, yapf, isort, black all pass).

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request adds a pool health summary to the job waiting spinner in sky/jobs/utils.py, providing users with visibility into worker statuses (e.g., ready, failed, provisioning) for pending jobs. Feedback suggests refactoring the _get_pool_health_summary function to use a single-pass iteration and collections.Counter for better efficiency and deterministic output.

[cblmemo]

According to the scenario you described, should we show how many pool replica is occupied by other jobs as well?

[lloyd-brown]

According to the scenario you described, should we show how many pool replica is occupied by other jobs as well?

Ah yes I like that, will work on adding that!