Slice firmware platform: bread-slice-core and the firmware spec #4

DCMT and RLHT share board code by copying it, so every fix to the e-stop, watchdog, status LED or boot path has landed as a pair of PRs. This campaign:

1. defines the shared behaviour in the BREADS standard;
2. puts one implementation of it in bread-slice-core;
3. moves both Slices onto it, with each move proven by host tests.

Decision and rationale: #2.

1. Standard (feastorg/bread-infra)

• spec/firmware-layering.md: the layers (Slice firmware, core, contracts, CRUMBS, device libraries, standard), what may depend on what, the admission test for core, and how code moves into it.
• spec/slice/firmware.md: status LED states, e-stop sampling and boot behaviour, the watchdog latch, and serial console rules. Open points to settle in that PR:
  • e-stop polarity: the standard says an open chain means stop, while the firmware asserts on LOW with a pull-up;
  • whether faults get their own LED colour;
  • which outputs need defined levels from reset.

2. Library (new repo feastorg/bread-slice-core)

• Repo, MPL-2.0, and CI (host tests plus AVR builds for 328P and 4809): feat: add the e-stop, watchdog, status LED and parsing modules bread-slice-core#1.
• Registry publishing on tag. 0.1.0 was published by hand.
• v0.1.0 modules, each with host tests (feat: add the e-stop, watchdog, status LED and parsing modules bread-slice-core#1, released as v0.1.0):
  • status LED (from refactor(firmware): drive the status LED without FastLED Slice_DCMT#55);
  • e-stop sampling, debounce, count and boot wait;
  • watchdog timer and latch;
  • strict number parsing.

3. Migration (one PR per Slice, each passing that Slice's host tests unchanged)

• Add a host test suite that builds the real firmware against CRUMBS Slice_DCMT#58 Host test suite for DCMT. This gates the DCMT migration.
• test(firmware): add host tests that build the firmware against CRUMBS Slice_RLHT#56 Host test suite for RLHT. This gates the RLHT migration.
• refactor(firmware): move e-stop, watchdog and LED onto bread-slice-core Slice_DCMT#77 DCMT adopts bread-slice-core, and its local copies go. refactor(firmware): drive the status LED without FastLED Slice_DCMT#55 is closed as replaced.
• refactor(firmware): move e-stop, watchdog and LED onto bread-slice-core Slice_RLHT#57 RLHT adopts bread-slice-core, and FastLED goes.

In both migrations test_status_led changed to read the core's LED frame in place of the FastLED stub; the other tests kept their checks. Neither migration has run on a board yet.

Related, tracked elsewhere

• Route serial and CRUMBS through one validated command path, behind host tests #3, the one-command-path epic: its strict parsing child is written in core; validate-and-apply stays per Slice.
• [Proposal] Define a common persistent-configuration contract for Slices bread-crumbs-contracts#61 (persistence): its record format and non-blocking writer become a core module once Route serial and CRUMBS through one validated command path, behind host tests #3's blockers clear.
• Automate PlatformIO dependency updates (Renovate?) #1 (Renovate): there will be a third shared library to keep pinned.
• is_command_op() duplication: goes to CRUMBS as a has-handler query, not to core.
• Safety epics Closed-loop modes can drive a motor further or faster than commanded Slice_DCMT#59 and Nothing bounds heater temperature but the setpoint and a working thermocouple Slice_RLHT#49 don't wait on this campaign.

Done when

• both specs are merged in bread-infra;
• bread-slice-core 0.1.x is published, with CI green on host tests and both AVR targets;
• neither Slice contains its own copy of a core module;
• Where shared Slice board code lives (status LED driver first) #2 is closed.