feat(mcp-server): advanced tools — wait/analyze/suggest/estimate (Phase 2)

Add four tools closing the submit→watch→analyze→suggest loop in the
request/response MCP model:

- scheduler_wait_job: poll a job to a terminal state or timeout.
- scheduler_analyze_failure: gather job record + output logs for failure
  analysis (interpretation left to the LLM).
- scheduler_suggest_resources: conservative processor/node suggestions from
  current cluster availability.
- scheduler_estimate_wait_time: rough wait estimate from pending count and
  free cores (guidance, not a guarantee).

All verified live on t1 against the running apiserver.

Co-Authored-By: Claude <noreply@anthropic.com>

src/frontend/mcp-server/cmd/main.go

@@ -112,6 +112,53 @@ func main() {
 		mcp.WithDestructiveHintAnnotation(false),
 	), app.HandleClusterMetrics)
 
+	// ── advanced tools (Phase 2) ──
+	// These help the agent close the submit → watch → analyze → suggest loop
+	// against the request/response MCP model.
+
+	// ── scheduler_wait_job ──
+	s.AddTool(mcp.NewTool("scheduler_wait_job",
+		mcp.WithDescription("Poll a job until it reaches a terminal state (DONE/EXIT) or a timeout. Returns the final status and elapsed time. Use after scheduler_submit to follow a job to completion."),
+		mcp.WithString("id",
+			mcp.Description("Job ID to wait for"),
+			mcp.Required(),
+		),
+		mcp.WithNumber("timeout_s", mcp.Description("Maximum seconds to wait (default 120)")),
+		mcp.WithNumber("poll_interval_s", mcp.Description("Poll interval seconds (default 3)")),
+		mcp.WithReadOnlyHintAnnotation(true),
+		mcp.WithDestructiveHintAnnotation(false),
+	), app.HandleWaitJob)
+
+	// ── scheduler_analyze_failure ──
+	s.AddTool(mcp.NewTool("scheduler_analyze_failure",
+		mcp.WithDescription("Gather everything about a failed (EXIT) job — job info, execution hosts, and output logs — for failure analysis. Returns structured data; interpret the likely cause from the logs and job metadata."),
+		mcp.WithString("id",
+			mcp.Description("Job ID of the failed job"),
+			mcp.Required(),
+		),
+		mcp.WithNumber("log_tail", mcp.Description("Max log lines to include (default 200)")),
+		mcp.WithReadOnlyHintAnnotation(true),
+		mcp.WithDestructiveHintAnnotation(false),
+	), app.HandleAnalyzeFailure)
+
+	// ── scheduler_suggest_resources ──
+	s.AddTool(mcp.NewTool("scheduler_suggest_resources",
+		mcp.WithDescription("Suggest resource parameters (processors, walltime, memory) for a job based on current node capacity and queue state. Returns concrete recommendations for scheduler_submit."),
+		mcp.WithString("command", mcp.Description("The command the job will run (used to gauge footprint)")),
+		mcp.WithString("queue", mcp.Description("Target queue (default normal)")),
+		mcp.WithReadOnlyHintAnnotation(true),
+		mcp.WithDestructiveHintAnnotation(false),
+	), app.HandleSuggestResources)
+
+	// ── scheduler_estimate_wait_time ──
+	s.AddTool(mcp.NewTool("scheduler_estimate_wait_time",
+		mcp.WithDescription("Estimate how long a pending job would wait before dispatch, based on current pending jobs, node capacity, and queue state. Returns a rough estimate — treat as guidance, not a guarantee."),
+		mcp.WithString("queue", mcp.Description("Target queue (default normal)")),
+		mcp.WithNumber("processors", mcp.Description("Number of processors the job requests (default 1)")),
+		mcp.WithReadOnlyHintAnnotation(true),
+		mcp.WithDestructiveHintAnnotation(false),
+	), app.HandleEstimateWaitTime)
+
 	// ── write tools (Phase 1) ──
 	// Write tools are annotated destructive so the MCP client (e.g. Claude
 	// Code) prompts the user for confirmation before executing them — the