Created by: FrostML

Fix the seqconv_eltadd_relu fault during cpu multi-threads inference.

Before this PR, the seqconv_eltadd_relu_fuse_pass will randomly or occasionally cause a seg-fault during cpu multi-thread inference with cloned predictors.

This PR fix this bug by removing the "ColMat" in AddOuput().AsIntermediate() and providing an intermediate tensor during kernel computing.