ci(nrf54l15,nrf52840): drop dead-fork PR override, build at MicroPython v1.27.0 pin

[cumin777]

Problem (ref #28)

Both workflows overrode the lib/micropython submodule before building:

git reset --hard origin/master
gh pr checkout 18030

Upstream micropython PR #18030 (zephyr: supports adding external user modules) was closed unmerged on 2025-11-20 and its source fork has since been deleted, so gh pr checkout fails with upload-pack: not our ref and CI has been red for both boards.

Why deleting the override is safe for these two boards

1. PR #18030's entire diff is one file, three additive lines (ports/zephyr/CMakeLists.txt: include(usermod.cmake), into QSTR, into include dirs). Neither workflow passes -DUSER_C_MODULES — grep across both files has zero matches — so those lines were a no-op for these boards. There is nothing else the fork could have contributed; the diff is the complete set.
2. The last-green base is a strict ancestor of v1.27.0. The PR head tree was fork-base 9939565 (master, 2025-08-26) + 3 lines; compare v1.27.0...9939565 → ahead_by=0, behind_by=409. So v1.27.0 strictly contains everything the green builds had — no feature can be lost by the move.
3. The repo is already pinned there. lib/micropython gitlink = 78ff170 = the v1.27.0 tag commit, and the green nRF54LM20A/B jobs build at exactly that submodule pin. This PR makes nRF54L15/nRF52840 do the same instead of moving the submodule to a deleted fork.

Scope (single variable changed)

• ❌ removed: reset --hard origin/master + gh pr checkout 18030
• ✅ build at the repo-wide pin (v1.27.0 via actions/checkout submodules, logged with git describe)
• untouched: NCS sdk-nrf v3.0.2, Zephyr SDK 0.17.0, the banner.c version.h shim, board overlays (BOARD_ROOT stays this repo)

What CI adjudicates

v1.27.0 (409 commits newer than the 2025-08-26 snapshot) × NCS v3.0.2 is the one unproven combination — that is exactly what this trial run tests. Fallback if red: pin the submodule checkout to 9939565 (the exact last-green base, already used as MP_BASE in tools/xiao_stm32c5/build.sh) and treat v1.27.0 unification as a follow-up with a compat patch, mirroring the micropython/patches/v1.27.0/ pattern.

Closes half of the "4 red boards" in #28 (MG24 needs USER_C_MODULES — patch or native usermod; RA4M1 needs #16409).

🤖 Generated with Claude Code

[cumin777]

Environment change notes (pre-merge summary)

What these boards were building on before (the "old environment"):

Both workflows overrode the lib/micropython submodule before compiling:

git reset --hard origin/master
gh pr checkout 18030     # upstream micropython PR, "zephyr: supports adding external user modules"

• PR #18030 was closed unmerged (author-closed 2025-11-20), and its source fork has since been deleted — GitHub can no longer serve the PR head ref, so gh pr checkout fails with upload-pack: not our ref. That is the entire reason CI went red for both boards (~2026-07). Nothing on the Seeed side changed.
• The PR's full diff is 3 additive lines in ports/zephyr/CMakeLists.txt (USER_C_MODULES plumbing). Neither board passes -DUSER_C_MODULES, so it was functionally a no-op here. What the boards were effectively compiling was the fork's base: master commit 9939565 (2025-08-26) — a frozen snapshot, not a release, and no longer fetchable once the fork died.

What they build on now:

The repo-wide pin — lib/micropython at v1.27.0 (78ff170) — the same model as the green nRF54LM20A/B jobs. 9939565 is a strict ancestor of v1.27.0 (compare v1.27.0...9939565 → behind_by 409, ahead_by 0), so nothing from the old base is lost. NCS v3.0.2, Zephyr SDK 0.17.0 and the board overlays are untouched.

Test status / caveat:

✅ CI compiles both firmwares green (artifacts: zephyr.elf/.hex for nRF54L15, zephyr.elf/.uf2 for nRF52840).

⚠️ No on-hardware validation has been done yet. The firmware builds but has not been flashed to or smoke-tested on an actual XIAO nRF54L15 / nRF52840 (REPL over USB, basic imports, GPIO/BLE presence). Recommend a hardware smoke test before cutting a release that ships these images.