P4: checkpoint battery — durable-resume seam + run.Checkpointer handle

Plugs into run.Ports.Checkpointer (the executor's call site is a P2 follow-up;
this provides the seam + impls ahead of it):
- checkpoint.go: CheckpointStore seam + RunCheckpoint{Meta, Messages, Iteration,
  ActivePhase} + RunCheckpointMeta (mirrors mort's agentexec types).
- handle.go: New(store, meta, throttle, now) -> run.Checkpointer. Save writes a
  throttled snapshot; Complete/Fail delete it (a cleanly finished or terminally
  failed run is NOT a recovery candidate; a shutdown-interrupted run never calls
  them, so its checkpoint survives ListInterrupted at boot). nil store -> no-op.
- memory.go: NewMemory() default (with the honest caveat that in-memory does
  not survive the restart it exists to recover from — a durable store is mort's).

Tests: save+complete clears the recovery candidate; throttle skips in-window
saves; nil-store is a clean no-op. Core imports ZERO from checkpoint.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

checkpoint/checkpoint.go

@@ -0,0 +1,50 @@
+// Package checkpoint is the durable-resume battery: it persists a run's
+// resumable progress so a run interrupted by a shutdown can be recovered and
+// continued on the next boot, rather than silently lost. It plugs into
+// run.Ports.Checkpointer.
+//
+// Mort backs CheckpointStore with its durable-job table; Memory() is the
+// zero-dependency default; contrib/store can add a SQLite one. NOTE: the
+// executor's call into run.Ports.Checkpointer is a P2 follow-up — this battery
+// provides the seam + impls ahead of that wiring.
+package checkpoint
+
+import (
+	"context"
+	"time"
+
+	"gitea.stevedudenhoeffer.com/steve/majordomo/llm"
+)
+
+// RunCheckpointMeta is the run attribution needed to resume a run from scratch
+// (mirrors mort's agentexec.RunCheckpointMeta).
+type RunCheckpointMeta struct {
+	RunID       string
+	AgentID     string
+	AgentName   string
+	CallerID    string
+	ChannelID   string
+	GuildID     string
+	Prompt      string
+	ModelTier   string
+	ParentRunID string
+}
+
+// RunCheckpoint is one persisted snapshot of a run's resumable progress.
+type RunCheckpoint struct {
+	Meta        RunCheckpointMeta
+	Messages    []llm.Message // conversation so far
+	Iteration   int           // completed agent-loop iterations
+	ActivePhase string        // current phase name (multi-phase agents); "" otherwise
+	UpdatedAt   time.Time
+}
+
+// CheckpointStore persists run checkpoints keyed by run id. A live checkpoint
+// means "this run was in flight and not cleanly finished"; Complete/Fail delete
+// it. ListInterrupted returns every surviving checkpoint at boot for recovery.
+type CheckpointStore interface {
+	Save(ctx context.Context, cp RunCheckpoint) error
+	Load(ctx context.Context, runID string) (*RunCheckpoint, error)
+	Delete(ctx context.Context, runID string) error
+	ListInterrupted(ctx context.Context) ([]RunCheckpoint, error)
+}