test: migrate Fam4 static-init/DATA-emit value tests to @test (#30)

Continues the test-arch tower past Fam8-13. 11 module-level static-init
/ DATA-emit value drivers move from test/wcc/*_run.c into @test row-
tables under test/lang/; every classification empirically re-probed at
HEAD (refuting two stale worklist tags).

- value rows -> test/lang/*_test.ww (11 files)
- reject rows -> runww //ww:error carriers (3, dual-stage non-vacuous;
  947 const-divzero confirmed a both-stage compile-reject, not run-exit)
- 840_zeroinit, 944_array_zeroinit, 989_arrlit_tail_zero kept as byte-id
  .c pins (zero-over-dirtied-frame / DATAW-length is byte-id-blind to a
  runtime @test; #263), mutation-gated
- repoint two stale comment refs to deleted test names (719,
  989_structlocal_frame)

Migrated static-init @test ride the cs==ww T2 byte-id gate, preserving
DATA-emit byte-id. 2D global-struct array-field read (#137/#150)
confirmed cs==ww + correct at HEAD. Coverage parity verified row-by-row;
two-round reviewed. Floor ratchet follows.

test/wcc/719_signed_data_emit.c

@@ -11,8 +11,9 @@
  *     negative literal should have produced 0xFF... (link succeeded;
  *     runtime read 0). The array arm hit the same gap.
  *
- * Both 923_signed_data_emit_run and the broader bootstrap byte-id
- * (995_self_rebuild) would catch a future regression, but this row
+ * Both test/lang/signed_data_emit_test.ww (the migrated runtime rows)
+ * and the broader bootstrap byte-id (995_self_rebuild) would catch a
+ * future regression, but this row
  * pins the *asm shape* itself — a future cgen refactor that emits
  * the slot via a different directive (e.g. via DATA + DATAR rather
  * than DATAW) would silently divergent even with green semantics.