chore: 初始开源发布

iotlcm-worker-svr：异步执行中枢（RabbitMQ 消费者 + 全量同步编排）。

- Go module：iotlcm-worker-svr → git.5iot.com/iotlcm-public/iotlcm-worker-svr（214 文件）
- 凭据与内部标识脱敏：MySQL/Redis 凭据、私网与生产 IP、内网域名、
  库名 xiujie_admin → iotlcm、哨兵名占位
- 159 处真实手机号 → 合法 11 位编造号（保留号段与后 4 位，维持互异，
  不破坏「校验 MSISDN 格式」等测试断言）
- 移除内部资产：CI 触发器、Jenkins 触发测试文档、预演门禁与镜像
  digest 记录
- Dockerfile 改用公共基础镜像（golang:1.25-alpine → alpine:3.23），
  移除写死的 git 凭据
- README 重写（架构/全量同步模型/配置/已知失败），补 Apache-2.0
  LICENSE，LF 行尾 + .gitattributes

已知：3 个测试包在源仓同一提交上同样失败（client/upstream、
service/allocate、service/inbound），属既有缺陷，本仓未引入亦未修复。

internal/fullsync/contract/interfaces.go

@@ -0,0 +1,145 @@
+package contract
+
+import (
+	"context"
+	"time"
+)
+
+// RunReader is the HTTP lookup boundary. Fall back to archive only when the
+// live reader returns ErrNotFound, not on transient live storage failures.
+type RunReader interface {
+	GetRun(ctx context.Context, runID string) (RunView, error)
+}
+
+type RunListFilter struct {
+	SyncKind SyncKind
+	Limit    int
+}
+
+type RunList struct {
+	Items []RunView `json:"items"`
+	Total int       `json:"total"`
+}
+
+type RunLister interface {
+	ListRuns(ctx context.Context, filter RunListFilter) (RunList, error)
+}
+
+type LiveRunReader interface {
+	GetRunMeta(ctx context.Context, runID string) (RunMeta, error)
+}
+
+type ArchivedRunReader interface {
+	GetArchivedRun(ctx context.Context, runID string) (RunArchive, error)
+}
+
+type RoundStatsReader interface {
+	GetRoundStats(ctx context.Context, runID string) ([]RoundStats, error)
+}
+
+// FailurePager returns live details only. An archived summary without live
+// details returns ErrFailureDetailsUnavailable, never a misleading empty page.
+// Items must serialize as [] on empty pages; NextCursor alone signals EOF.
+type FailurePager interface {
+	ListFailures(ctx context.Context, runID string, page PageRequest) (FailurePage, error)
+}
+
+type ReplayRequester interface {
+	RequestReplay(ctx context.Context, request ReplayRequest) (ReplayAccepted, error)
+}
+
+type ActionReader interface {
+	GetReplayAction(ctx context.Context, actionID string) (ReplayAction, error)
+}
+
+// ArchiveWorkDiscovery returns pending/retryable work from a durable index.
+// Terminal run registration must not depend on an in-memory notification.
+type ArchiveWorkDiscovery interface {
+	ListArchiveWork(ctx context.Context, page PageRequest) (ArchiveWorkPage, error)
+}
+
+// RegisterArchiveWork idempotently schedules a terminal run/version. Since the
+// index can occupy another Redis slot, callers must reconcile registration
+// after a crash between completing the run and updating the index.
+type ArchiveWorkRegistrar interface {
+	RegisterArchiveWork(ctx context.Context, work ArchiveWork) error
+}
+
+type ArchiveWorkLoader interface {
+	LoadArchiveWork(ctx context.Context, work ArchiveWork) (RunArchive, error)
+}
+
+// ArchiveRegistrar conditionally records a durable publication against the
+// work version. Repeating the same receipt is a no-op; a different version or
+// conflicting receipt returns ErrConflict. Failure retains discoverable work.
+type ArchiveRegistrar interface {
+	RegisterArchive(ctx context.Context, work ArchiveWork, receipt ArchiveReceipt) error
+	RegisterArchiveFailure(ctx context.Context, work ArchiveWork, message string) error
+}
+
+// ReplayActionCreator atomically reserves (sourceRunId, idempotencyKey) with
+// its fingerprint and action. Same key/request returns the original action
+// regardless of candidate IDs; different request returns ErrConflict. The
+// candidate starts accepted and fixes the target ID before any target writes.
+type ReplayActionCreator interface {
+	GetOrCreateReplayAction(ctx context.Context, request ReplayRequest, candidate ReplayAction) (ReplayActionCreation, error)
+}
+
+// Discovery includes accepted, processing, cleanup_pending and receipt_pending
+// actions after crashes. Completed/needs_attention actions require no automatic
+// work. Lease acquisition, not listing, chooses an exclusive owner.
+type ReplayWorkDiscovery interface {
+	ListReplayWork(ctx context.Context, page PageRequest) (ReplayActionPage, error)
+}
+
+type ReplayLeaseStore interface {
+	AcquireReplayLease(ctx context.Context, actionID, owner string, ttl time.Duration) (ReplayLease, error)
+	RenewReplayLease(ctx context.Context, lease ReplayLease, ttl time.Duration) (ReplayLease, error)
+	ReleaseReplayLease(ctx context.Context, lease ReplayLease) error
+}
+
+// All updates are fenced and compare ExpectedRevision. Expired/replaced leases
+// return ErrLeaseLost; revision conflicts return ErrConflict. Implementations
+// reject progress regression and preserve immutable action identity.
+type ReplayProgressStore interface {
+	UpdateReplayAction(ctx context.Context, update ReplayActionUpdate) (ReplayAction, error)
+}
+
+// The implementation must preserve scan continuity while source entries are
+// cleaned. This can use a durable selection/snapshot; deleting records must not
+// make a later cursor skip an unprocessed failure. Tokens are bound to actionID.
+type ReplayFailurePager interface {
+	ListReplayFailures(ctx context.Context, actionID string, page PageRequest) (FailureRecordPage, error)
+}
+
+// TargetRunCreator is idempotent: retrying the same target preserves counters;
+// incompatible source/operation/month/cycle returns ErrConflict. Replay targets
+// are distinct runs and must not mutate the source's historical FailedCount.
+type TargetRunCreator interface {
+	EnsureTargetRun(ctx context.Context, meta RunMeta) (RunMeta, error)
+}
+
+// TargetEnqueueResult distinguishes a durable duplicate from a rejected write.
+// A zero NewlyEnqueued count alone is never sufficient evidence for cleanup.
+// Every input TaskID must be durably present on nil error; otherwise return an
+// error and let the caller safely retry the identical batch. Both counts are
+// unique task counts, and their sum equals the number of distinct input IDs.
+type TargetEnqueueResult struct {
+	NewlyEnqueued  int64
+	AlreadyPresent int64
+}
+
+type TargetRunEnqueuer interface {
+	// Tasks must match the target run and channel and use stable logical IDs.
+	EnqueueTargetBatch(ctx context.Context, lease ReplayLease, runID, chanKey string, tasks []Task, nextExecuteAt time.Time) (TargetEnqueueResult, error)
+	MarkTargetProducerDone(ctx context.Context, lease ReplayLease, runID string) error
+}
+
+// Cleanup is idempotent and conditional on the lease, exact source versions,
+// and durable evidence of target acceptance. It changes pending failures only,
+// never historical failedCount. Conflicts must remain visible for attention.
+// Source keys and action/target keys may occupy different Redis Cluster slots;
+// implementations need recoverable receipts, not cross-slot Lua assumptions.
+type ReplayFailureCleaner interface {
+	CleanupReplayFailures(ctx context.Context, cleanup ReplayCleanup) (ReplayCleanupResult, error)
+}