feat(auth): add PAT auth, scoped REST/MCP access, CSRF, and tenant isolation

Introduce Personal Access Tokens (nzp_*) as a stateless auth path alongside
JWT, gated per-endpoint by a scope middleware (nezha:{resource}:{verb}) with
fail-closed empty-scope defaults and a server-id whitelist. Self-management
endpoints (profile, api-tokens, oauth2 bind, refresh-token) explicitly reject
PATs to block privilege-escalation chains. A revoke registry tears down active
long-lived connections (terminal, fm, ws, transfer, mcp) the moment a PAT is
deleted, with a tombstone closing the revoke->register race.

Add an MCP endpoint that proxies tool calls (exec, fs read/write/delete,
transfer) to agents over gRPC, guarded by origin/DNS-rebinding checks, a
per-token rate limiter, audit logging, and a kill switch. Serialize all
sends through the IOStream wrapper to honour grpc-go's concurrency contract.

Add CSRF double-submit protection on unsafe cookie-authenticated methods,
exempting authenticated PAT requests by context identity (not a forgeable
Authorization header). Apply visibility/whitelist filtering consistently
across list, get-by-id, and mutate paths to enforce tenant isolation.

Migrate legacy mcp:* scopes: rewrite read/exec to nezha:* equivalents and
drop dangerous write/delete/wildcard grants.

Co-authored-by: cloudcode <cloudcode@users.noreply.github.com>

service/rpc/io_stream_race_test.go

@@ -0,0 +1,92 @@
+package rpc
+
+import (
+	"io"
+	"sync"
+	"testing"
+	"time"
+)
+
+// nopRWC is a minimal io.ReadWriteCloser used for race tests; Close is a
+// no-op so the racer goroutines do not panic on shared state.
+type nopRWC struct{}
+
+func (nopRWC) Read(p []byte) (int, error)  { return 0, io.EOF }
+func (nopRWC) Write(p []byte) (int, error) { return len(p), nil }
+func (nopRWC) Close() error                { return nil }
+
+// H10 regression: UserConnected/AgentConnected mutate stream.userIo /
+// stream.agentIo without holding ioStreamMutex, while WaitForAgent /
+// RevokeStreamsForPurpose / RevokeStreamsForServer read & close the same
+// fields under the lock. The go race detector catches it deterministically
+// under -race; without the fix this test fails.
+func TestIOStream_AgentConnectedIsRaceFreeUnderLock(t *testing.T) {
+	h := NewNezhaHandler()
+	const streamId = "race-test"
+	h.CreateStream(streamId, 1, 1)
+	t.Cleanup(func() { _ = h.CloseStream(streamId) })
+
+	var wg sync.WaitGroup
+	wg.Add(3)
+
+	go func() {
+		defer wg.Done()
+		// repeatedly attach an agent
+		for i := 0; i < 200; i++ {
+			_ = h.AgentConnected(streamId, nopRWC{})
+		}
+	}()
+
+	go func() {
+		defer wg.Done()
+		// concurrently attach a user
+		for i := 0; i < 200; i++ {
+			_ = h.UserConnected(streamId, nopRWC{})
+		}
+	}()
+
+	go func() {
+		defer wg.Done()
+		// Revoker takes the write lock and reads the same userIo/agentIo
+		// fields the unsynchronised writers above are setting. Use a real
+		// targetServerID (1) so RevokeStreamsForServer actually inspects
+		// the entry's userIo/agentIo before deleting.
+		for i := 0; i < 200; i++ {
+			h.RevokeStreamsForServer(1)
+			h.CreateStream(streamId, 1, 1)
+		}
+	}()
+
+	wg.Wait()
+}
+
+// StartStream reads stream.userIo/agentIo while it waits for both endpoints.
+// Those reads must be lock-protected against the concurrent writes done by
+// UserConnected/AgentConnected; otherwise -race flags the data race on the
+// interface fields.
+func TestIOStream_StartStreamReadsAreRaceFree(t *testing.T) {
+	h := NewNezhaHandler()
+	const streamId = "startstream-race"
+	h.CreateStream(streamId, 2, 2)
+	t.Cleanup(func() { _ = h.CloseStream(streamId) })
+
+	var wg sync.WaitGroup
+	wg.Add(3)
+
+	go func() {
+		defer wg.Done()
+		_ = h.StartStream(streamId, 50*time.Millisecond)
+	}()
+	go func() {
+		defer wg.Done()
+		time.Sleep(5 * time.Millisecond)
+		_ = h.AgentConnected(streamId, nopRWC{})
+	}()
+	go func() {
+		defer wg.Done()
+		time.Sleep(5 * time.Millisecond)
+		_ = h.UserConnected(streamId, nopRWC{})
+	}()
+
+	wg.Wait()
+}