Fix duplicate subnet allocations

[robmry]

- What I did

Keep allocated subnets in-order, so that they're not mistakenly reallocated due to a gap in the list where misplaced subnets should have been.

• fixes Docker-27.0.2: "invalid pool request: Pool overlaps with other one on this address space" #48069
• fixes Overlay networks get automatically created in overlapping segments docker/cli#5204
• fixes [Bug]: cannot create network ... : conflicts with network ... : networks have overlapping IPv4 testcontainers/testcontainers-java#8813
• introduced in libnet/ipams/default: introduce a linear allocator #47768

- How I did it

The iterator over allocated subnets was incremented too early, this change moves it past three clauses in addrSpace.allocatePredefinedPool().

- How to verify it

The three new unit tests correspond to a separate failure caused by incrementing before each of them.

Without the fix, the three new tests fail like this ...

=== FAIL: libnetwork/ipams/defaultipam TestPoolAllocateAndRelease/allocate_after_reserved (0.00s)
    address_space_test.go:475: assertion failed: reserved.Overlaps(subnet) is true: subnet 10.0.1.0/24 allocated in reserved range 10.0.0.0/16

=== FAIL: libnetwork/ipams/defaultipam TestPoolAllocateAndRelease/reallocate_first_subnet (0.00s)
    address_space_test.go:472: assertion failed: exists is true: subnet 10.0.0.0/24 allocated to n4, reallocated for n5

=== FAIL: libnetwork/ipams/defaultipam TestPoolAllocateAndRelease/reallocate_after_release (0.00s)
    address_space_test.go:472: assertion failed: exists is true: subnet 10.0.1.0/24 allocated to n5, reallocated for n6

As a follow-up, it might be good to add something to the tests to assert that allocated is ordered (and maybe to the production code to log an error if the ordering breaks).

- Description for the changelog

Fix a regression that caused duplicate subnet allocations when creating networks.

[thaJeztah]

LGTM