libsel4vmmplatsupport: use vm->entry for zImage also

[axel-h]

This is a follow-up from the change introduced by https://github.com/seL4/seL4_projects_libs/pull/81

• Remove hard coded fall-back address for zImage also.
• inline zImage_get_load_address(), so things get a bit simpler .
• Add comment about zImage usage in Linux. Seems it's 32-bit only anyway.

[axel-h]

To me, the vm->entry usage seems quite strange. Loading behavior of the kernel depends on the image type:

• for a binary, the parameter load_base_addr is ignores and the blob is loaded vm->entry. That seem the only way how things work in AARCH64
• a zImage (which practically exists for AARCH32 only), load_base_addr is also ignored and the address from the header is used. Or, if this is zero, it falls back to vm->entry with this PR (before it was indeed load_base_addr + offset then.),

So it seems vm->entry is slightly misused in what we have currently? Looks like there should be a vm->kernel_addr also that must be given in case the kernel is a raw binary.

[axel-h]

With the changes of this PR it seems we don't really need 'vm.entry' that 58dc2820 recently added (https://github.com/seL4/seL4_projects_libs/pull/81). We should instead pass the value in load_base_addr, as this parameter has been practically unused. That would nicely simplify things then.