build: use `NVME_` prefix for all `HAVE_FOO` defines

[igaw]

Update all HAVE_ defines using a NVME_HAVE_ prefix to avoid clashes with other projects defining their own HAVE_ defines, e.g. the Python binding uses HAVE_MMAP too.

[martin-belanger]

I had a similar issue with fallthrough which is also defined in the Python bindings (swig). I was able to fix it by using an #undef, but I think your idea is better.

Ref: https://github.com/linux-nvme/nvme-cli/blob/master/libnvme/libnvme/nvme.i#L12

[igaw]

I'd like to keep the falltrough for the C code. But the rest is fair game to me

[martin-belanger]

IIRC, SWIG redefines fallthrough exactly the same as ours. So, #undef'ing ours does not break anything. The C code will just see SWIG's definition.

However, maybe we could rename our fallthrough to nvme_fallthrough or NVME_FALLTHROUGH. What do you think?

[igaw]

I would like to keep in our *.c files the fallthrough as name. I am fine with the undef. Yes it's a bit hacky but pretty much contained to the python binding. SWIG could test for the existence of fallthrough.

We have a bunch of those in nvme-cli but not only one user in libnvme. We could try to get rid of this one and avoid defining fallthrough for the library. Or we just defined compiler-attributes.h and remove it from the global scope.

Anyway, my point is nvme_fallthrough looks wrong to my eyes. It's a compiler annotation everyone understand but nvme_fallthrough is what?