convolution in the new MLIR bridge to improve MXU efficiency of low batch size
convolutions._

## Background

For image models, the first layer is usually not MXU friendly as it has a
feature size of 3. This results in poor performance especially with small batch.

One way to address this issue is to use the `space-to-depth` transform. This
optimization tiles the 2x2 space dimensions to the feature dimension so that the

	//
	// TODO(austin,mknyszek): This still has bad worst-case
	// throughput. For example, this could find just one free slot
	// on the 100th swept span. That limits allocation latency, but
	// still has very poor throughput. We could instead keep a
	// running free-to-used budget and switch to fresh span
	// allocation if the budget runs low.
	spanBudget := 100

	var s *mspan
	var sl sweepLocker