R6(a): de-stage code, docs, READMEs; split stage6_test

Mechanical, behaviour-preserving removal of Stage N / TODO-N / phase (RN)
references from comments, doc-comments, service READMEs, the current-state docs
(ARCHITECTURE, FUNCTIONAL+_ru, TESTING, UI_DESIGN), config-file comments, and the
.fbs/.proto schema comments. PLAN.md / PRERELEASE.md / CLAUDE.md keep the stage
history.

- Rename the only stage-named identifiers: registerStage8 -> registerSocialOps,
  registerStage11 -> registerLinkOps (gateway transcode).
- Split stage6_test.go: TestEmailLoginFlow -> email_test.go,
  TestGuestAutoMatchLeavesNoStats (+ provisionGuest) -> account_test.go.
- Regenerated proto bindings (push.pb.go, telegram_grpc.pb.go) from the de-staged
  .proto comments; FB Go/TS bindings unchanged (flatc strips schema comments).

go build/vet/gofmt clean across modules; integration typecheck and pnpm check green.

loadtest/internal/scenario/scenario.go

@@ -2,7 +2,7 @@
 // assembles real games through the invitation flow, then runs each player's turn
 // loop (poll state, replay history, generate a legal move with the embedded solver,
 // submit it) plus a fraction of secondary operations. It exposes the moderate
-// realistic ramp agreed for the R2 early pass and a separate gateway-hammer.
+// realistic ramp and a separate gateway-hammer.
 package scenario
 
 import (
@@ -41,7 +41,7 @@ type RealisticConfig struct {
 	SecondaryProb  float64       // chance per tick of a non-move operation
 }
 
-// DefaultRealistic returns the moderate ramp agreed for the R2 early pass: 50 -> 200
+// DefaultRealistic returns the moderate ramp: 50 -> 200
 // -> 500 concurrent players, ~12 minutes per step, ~1 op/s per player.
 func DefaultRealistic() RealisticConfig {
 	return RealisticConfig{