Single-slot boot path is duplicated in ports; consider factoring it into `boot/bootutil`

[danielriege]

The single-slot (MCUBOOT_SINGLE_APPLICATION_SLOT) boot path exists twice, in two
ports, and not at all in the shared library:

File	Lines
boot/zephyr/single_loader.c	255
boot/mynewt/src/single_loader.c	136
boot/bootutil/src/	— none

Both provide boot_go() plus a local boot_image_validate(); Zephyr's
additionally provides boot_get_loader_state(), boot_get_image_max_sizes(),
boot_open_all_flash_areas(), boot_close_all_flash_areas() and
app_max_size(), of which the first four are only referenced by
boot_fetch_slot_state_sizes() in bootutil_misc.c, inside
#if defined(MCUBOOT_SERIAL_IMG_GRP_SLOT_INFO).

Neither implementation appears to depend on its host OS. In
boot/zephyr/single_loader.c I could find no Zephyr API call, no CONFIG_*
reference and no devicetree use; its includes are assert.h, bootutil/*,
bootutil_priv.h and mcuboot_config/mcuboot_config.h, and the only
non-bootutil symbols it references are the BOOT_LOG_* macros, which are
already part of the port configuration contract. The same appears to hold for
the Mynewt version.

Why it matters

boot_go() has two implementations — the multi-slot state machine in
boot/bootutil/src/loader.c, and the single-slot path — and the build system
selects between them (boot/zephyr/CMakeLists.txt, the
CONFIG_SINGLE_APPLICATION_SLOT branch). Because only the multi-slot
implementation lives in the shared library, an out-of-tree port that wants
single-slot support has to write the second one itself, even though it is
almost entirely calls into bootutil.

That also means the two in-tree copies can drift from each other and from
bootutil, and each has to be updated separately when the bootutil interface
changes.

Proposal

Move the OS-independent part of the single-slot boot path into
boot/bootutil/src/single_loader.c and have the Zephyr and Mynewt ports use it,
leaving anything genuinely port-specific behind. This would be a
behaviour-preserving refactor rather than a feature change, and would give
out-of-tree ports the same choice of boot_go() implementation that the in-tree
ports have.

Questions before attempting a PR

1. Is consolidating these two into bootutil a direction you would accept in
   principle, or is the duplication deliberate?
2. sim/Cargo.toml has features for overwrite-only, swap-move,
   swap-offset and so on, but none for single-slot, so as far as I can tell the
   single-slot path currently has no simulator coverage. For a
   behaviour-preserving move, would you want simulator coverage for single-slot
   added first (or as part of the same PR), or is the absence of coverage
   acceptable for a relocation?
3. If you would take it, do you have a preference on scope — move Zephyr's
   version and have Mynewt adopt it, or reduce to the smaller common subset and
   let each port keep what only it needs?

Happy to do the work if the direction is welcome; asking first per
docs/contributing.md.

[nordicjm]

I believe it was done as zephyr specific because of the lack of simulator testing, it would be good to get it all moved into common code and have tests for it (they would need to go together in the same PR). Zephyr's is the correct one and should be used