fix: Foundry E2E quick-wins (#111, #115, #116, #117)

[Dongbumlee]

Bundles the four lowest-risk findings from the Foundry agent end-to-end validation in #78 into a single, low-risk PR. No backend changes; all fixes are docs, templates, or CLI surface.

Fixes

• Closes Implement or hide 'agentops config validate' (currently a stub) #115 — implement agentops config validate -c <path>. Loads the given run.yaml through the same Pydantic models the runner uses, prints a one-line success summary or a clear error, and exits non-zero on failure.
• Closes Default avg_latency_seconds threshold (10s) is unrealistic for Foundry cloud eval #116 — raise default avg_latency_seconds thresholds across baseline bundles to match Foundry cloud-evaluation reality (15–25 s/row observed):
  • 10 s → 30 s for conversational_agent_baseline, model_quality_baseline, rag_quality_baseline, safe_agent_baseline
  • 15 s → 45 s for agent_workflow_baseline
  • Tutorial tables updated to match.
• Closes agent_workflow_baseline silently scores ~0 against tool-less agents #117 — document in docs/bundles.md that agent_workflow_baseline is only meaningful against agents that expose tool definitions, and that avg_latency_seconds measures the full pipeline rather than the agent alone.
• Closes Document agent_id resolution when version is omitted #111 — document in the Foundry agent tutorial and docs/how-it-works.md that omitting the :version suffix on a named Foundry agent resolves to latest at run time, and that CI / baseline runs should pin a version.

Out of scope (separate follow-up issues)

• Foundry cloud-eval: agent response missing from report.md #110 (empty Response in report.md), agentops eval compare: per-run column uses baseline's threshold for all runs #112 (compare uses baseline threshold for all runs), Foundry cloud-eval: per-row latency is faked (run-level avg duplicated) #113 (faked per-row latency) — these require backend / runner changes and are kept for separate review.

Validation

$ python3 -m pytest tests/ -x -q
285 passed in 1.27s

3 new unit tests for config validate (success, missing file, invalid schema).

Manual smoke:

$ agentops config validate -c .agentops/run-foundry-conversational.yaml
✅ ... is valid (version=1, target.type=agent, target.hosting=foundry, target.execution_mode=remote, bundle=conversational_agent_baseline, dataset=smoke-conversational)
$ agentops config validate -c does/not/exist.yaml; echo $?
Error: config file not found: does/not/exist.yaml
1