metal : fix MUL_MAT+ADD fusion when the residual is itself a MUL_MAT (llama/30100)

* metal : fix MUL_MAT+ADD fusion when the residual is itself a MUL_MAT

ggml_metal_op_mul_mat_mma picks the residual of a fused MUL_MAT+ADD as
"the ADD operand whose op is not MUL_MAT". When both operands of the ADD
are mat-mul outputs (x = W1 @ u + W2 @ v), that test is true for both, so
the residual resolves to the fused mat-mul's own, never-written output and
the kernel adds whatever that buffer holds.

The fusion check (ggml_metal_mul_mat_add_operand) already selects the
operand by identity; make the encoder do the same.

Clef decision models hit this in their head (proj_option_context @ ctx +
proj_option_lexical @ lex, 9 option rows): on Metal, /v1/systemone
probabilities collapse toward uniform (billing 0.28 where the CPU backend
gives 0.977, Cloudflare_clef-flash Q8_0), deterministic per memory layout,
correct with GGML_METAL_FUSION_DISABLE=1. Not a quantization issue: the
same file is right on CPU.

Add a MUL_MAT_ADD mode to test-backend-ops where the residual is a second
mat-mul; on Metal it fails 27 of 28 cases before this change (the one pass
is f16 n=2, under the MMA row threshold, so nothing fuses).

* Update tests/test-backend-ops.cpp

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

ggml/src/ggml-metal/ggml-metal-ops.cpp

@@ -2476,7 +2476,8 @@ static int ggml_metal_op_mul_mat_mma(ggml_metal_op_t ctx, int idx) {
 
     if (fuse_add) {
         dst = ctx->node(idx + n_fuse - 1);
-        res = dst->src[0]->op == GGML_OP_MUL_MAT ? dst->src[1] : dst->src[0];
+        // the residual is the other operand of the ADD, by identity: it can itself be a MUL_MAT output
+        res = dst->src[0] == op ? dst->src[1] : dst->src[0];
     }
 
     auto pipeline = ggml_metal_library_get_pipeline_mul_mv_mma_auto(lib, op, fuse_add);