mm: filemap: pass interleave index through filemap_alloc_folio

filemap_alloc_folio_noprof() hardcodes NO_INTERLEAVE_INDEX when
calling folio_alloc_mpol_noprof() for NUMA policy-based allocations.
This causes MPOL_INTERLEAVE to fall back to the task's global
il_prev counter instead of using the file offset for deterministic
page placement.

The only current user passing a non-NULL policy is
__filemap_get_folio_mpol(), called by KVM guest_memfd.  The page
index is already available at that call site but was never threaded
down to the allocator.

Add a pgoff_t ilx parameter to filemap_alloc_folio_noprof() and
pass it through to folio_alloc_mpol_noprof().  Update
__filemap_get_folio_mpol() to forward its index argument, and all
other callers (which pass NULL policy and never hit the mpol path)
to pass 0.

Fixes: 7f3779a3ac3e ("mm/filemap: Add NUMA mempolicy support to filemap_alloc_folio()")
Assisted-by: Claude:claude-opus-4-6
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>