deepctr 例子中 deep0 输入维度超过1024维后报错，降低维度后运行正常

仿造deepctr 的例子
```
reader.epochs(1).threads(1).batch_size(4).label_count(1)
emb1 = xdl.embedding('emb1', batch['sparse0'], xdl.TruncatedNormal(stddev=0.001), 8, 1024, vtype='hash')
emb2 = xdl.embedding('emb2', batch['sparse1'], xdl.TruncatedNormal(stddev=0.001), 8, 1024, vtype='hash')
reader.feature(name='deep0', type=xdl.features.dense, nvec=1025)
reader.startup()
```

运行上面代码时报错如下

`/home/yue.song/x-deeplearning/xdl/xdl/data_io/packer/pack_feature.cc:294] Check failed: blk->ts_[Block::kValue]->Shape()[1] == val.vector_size() dense feature=dense1 vector_size=`

已确认数据和代码的维度设置一致。
尝试把nvec 和对应的数据降低到1024维以内时，正常。

Start at xdl/data_io/packer/pack_feature.cc around line 294 and reproduce the failure using reader.feature with dense nvec=1025. Compare the behavior with the reported working case at 1024 dimensions and verify that the data and configured vector sizes remain consistent. Done means the reported dense-feature input no longer fails above 1024 dimensions.

Written by the indexing model from the issue text.