Bug-fixes

• Added GPU support for the syevd operator which ensures that there is GPU support for all linalg-operators.
• Bugfix for syevd on CPU such that it works for float32.
• Fixed API call when OMP_NUM_THREADS environment variable is set.
• Fixed MakeNonlossGradNode bug.
• Fixed bug related to passing dtype to array().
• Fixed some minor bugs for sparse distributed training.  
• Fixed a bug on Slice accessing uninitialized memory in param.begin in the file matrix_op-inl.h.
• Fixed gluon.data.RecordFileDataset.
• Fixed a bug that caused autograd to crash on some networks.