x86/hvm: track large memory mapped accesses by buffer offset

x86/hvm: track large memory mapped accesses by buffer offset

The code in hvmemul_do_io() that tracks large reads or writes, to avoid
re-issue of component I/O, is defeated by accesses across a page boundary
because it uses physical address. The code is also only relevant to memory
mapped I/O to or from a buffer.

This patch re-factors the code and moves it into hvmemul_phys_mmio_access()
where it is relevant and tracks using buffer offset rather than address.
Separate I/O emulations (of which there may be up to three per instruction)
are distinguished by linear address.

Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>