Support DeepSeek V4 Flash Vision Experimental

docs/issue100-vision/encoder.log

@@ -0,0 +1,11 @@
+
+running 1 test
+test engine::metal::deepseek_vision::tests::issue100_encoder_real_weights ... ds4: Metal device Apple M5 Max, 128.00 GiB RAM
+ds4: Metal 4 tensor API enabled for Tensor kernels
+ds4: drift-patch flags hc_stable=on norm_unify=on kv_raw_f32=off rope_exp2_log2=off math_safe=off tensor_matmul=on
+ds4: Metal model views created in 0.039 ms, residency requested in 6.538 ms, warmup 3.368 ms (mapped 889.64 MiB from offset 0.02 MiB)
+ds4: Metal mapped mmaped model as 1 overlapping shared buffers
+{"grid":[28,28],"aligned_grid":[10,10],"tokens":100,"width":224,"height":224,"values":409600}
+ok
+
+test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 563 filtered out; finished in 2.01s