Dont Support FusedBatchNorm OP

[theflyfish]

when applying blade on tensorflow 1.15，The whole graph is divided by fusedBatchNorm while all other op has been clusted into serval cluster,so i have some question:

1. why dont we support FusedBatchNorm op on MLIR mode, as in tf1.5 above and tf2.X，the fused batchnorm op is the default method for batchnorm； and if supported it , the whole graph will be clusted into one cluster which may be achieve better performance
2. if i want to add the FusedBatchNorm support ,what should i do
3. is the any other solution for me to fused the fusedbatchnorm into the cluster