#84 Add warm up iterations to NemoLite2D timing

[sergisiso]

Add multiple timing sections in the nemolite2D algorithm layer. One of them for warm up time-stepping iterations so device initalisation done in the first iteration don't contribute to the average iteration time.

Also adds HIP backend for kokkos and fixes some issues.

[sergisiso]

@arporter @rupertford This is not ready yet but I realised that changing nemolite2d.f90 is quite tedious because the file is replicated in all manual_implementations and in psykal. And it is important for fair comparison to have them all be equivalent. So I was thinking to do the same we do with namelist and create a original one and then the other be symbolic links to it. I am wondering where this original should be placed. Could you comment if you prefer:

• Keep it different files as they are now.
• Have a shared nemolite2d.f90 in manual_implementations root but not share with psykal
• Have a shared nemolite2d.f90 in the nemolite2d root folder
• Have a shared nemolite2d.f90 in the nemolite2d/common folder

[arporter]

I think having just one file is good and I'd prefer putting it in nemolite2d/common.

[arporter]

@sergisiso is this one ready for review?