Wrapper with tests for gemmBatchedEx!

[lpawela]

This introduces a wrapper for gemmBatchedEx!, allowing for batched mixed precision matrix multiplication. Based on the approach used for existing gemmEx! and gemm_batched!.

[maleadt]

Thanks! Couple of minor comments.

I'm not really fond of these interfaces, they're low-level and ugly, but we don't have anything better right now.
Do you plan to use these directly, or to extend e.g. NNlib.batched_mul?

[lpawela]

I'm not planning to add it NNlib, I'm using these directly. I will also add here (hopefuly today) the strided version.