Initial Oceananigans Kernel

[wsmoses]

This requries the fix for tracing as discussed on slack with @mofeing @jumerckx and others.

But assume that is resolved, and also llvm/llvm-project#121987 lands (cc @ftynse @ivanradanov @glwagner ) here is the IR.

ocean.mlir.txt

Per other conversation @ftynse we clearly need to figure out the SROA of those constants from the ABI shim (and inlining). Then from there we can start to do @ivanradanov's polyhedral optimizations.

[wsmoses]

Now with the various actually correct calling convention (and hopefully just one nv_sin)

ocean2.mlir.txt