Add 2D grid program_id semantics (ADR-0022)

tl.program_id(axis=0) returns local PE id within cube,
tl.program_id(axis=1) returns cube id. Enables cube-aware
sharding in benchmark kernels.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

tests/test_triton_emu.py

@@ -196,6 +196,23 @@ def test_tl_program_id():
     assert tl.num_programs(0) == 8
 
 
+def test_tl_program_id_axis1():
+    """axis=1 returns cube_id and num_cubes."""
+    tl = TLContext(pe_id=3, num_programs=8, cube_id=7, num_cubes=16)
+    assert tl.program_id(0) == 3
+    assert tl.program_id(1) == 7
+    assert tl.num_programs(0) == 8
+    assert tl.num_programs(1) == 16
+
+
+def test_tl_program_id_global():
+    """global_pid = cube_id * num_pes_per_cube + local_pe_id."""
+    pe_id, cube_id, num_pes = 5, 3, 8
+    tl = TLContext(pe_id=pe_id, num_programs=num_pes, cube_id=cube_id, num_cubes=16)
+    global_pid = tl.program_id(1) * tl.num_programs(0) + tl.program_id(0)
+    assert global_pid == cube_id * num_pes + pe_id
+
+
 # ── 12. tl.arange, tl.zeros, tl.full ─────────────────────────────