Remove our workspace size checking

[hobofan]

We originally wanted to remove the somewhat unneccesary needs_cudnn_workspace with #37 , but due to some reason always computing the neccesary workspace size via cudnn for bwd_data.

The current solution looks like this:

collenchyma-nn/src/frameworks/cuda/mod.rs

Lines 309 to 313 in d9bb5da

	// let workspace_size_bwd_data = API::get_convolution_backward_data_workspace_size(*CUDNN.id_c(), useable_algo_bwd_data.as_cudnn().unwrap(), *filter_desc.id_c(), *conv_desc.id_c(), *src_desc.id_c(), *dest_desc.id_c()).unwrap();
	let workspace_size_bwd_data = match try!(useable_algo_bwd_data.needs_cudnn_workspace()) {
	false => 1,
	true => API::get_convolution_backward_data_workspace_size(*CUDNN.id_c(), useable_algo_bwd_data.as_cudnn().unwrap(), *filter_desc.id_c(), *conv_desc.id_c(), *src_desc.id_c(), *dest_desc.id_c()).unwrap(),
	};

but optimally we would want to get rid of our own needs_cudnn_workspace heuristics.

[hobofan]

#40 introduced a fix for this and removed needs_cudnn_workspace.

However I had to introduce a workaround which is tracked #42.