Update libopeninv submodule to latest (jsphuebner 51a9e58)

[wjcloudy]

Summary

Bumps the libopeninv submodule from the currently-pinned 3a9c0594 to the current jsphuebner tip 51a9e58, and adapts the VCU. Hardware-validated (see Testing).

What the bump brings

libopeninv commits since 3a9c0594 include:

• 100 kbps (Baud100) CAN bit timings
• Indexed CAN-map transmit API + fix for the premature ~20-signal CAN-map limit
• SDO refactor — user-space SDO objects dispatched via CanSdo::ProcessSDO → SdoCommands

VCU changes in this PR

1. CanSdo — drop the GetPendingUserspaceSdo() main-loop poll. The new ProcessSDO already routes user-space SDO objects (serial 0x5000 / command 0x5002) to SdoCommands from HandleRx, so plain CanSdo suffices.
   • The SDO command path (incl. parm_save) runs in CAN-RX interrupt context — that's libopeninv's default ProcessSDO behaviour, not VCU code. Confirmed working on hardware (param save persists across reboot).
2. No Send() call-site changes — an earlier revision cast lengths to uint8_t to dodge an overload ambiguity; canhardware: remove ambiguous 2-arg Send overload (lets callers drop uint8_t casts) jsphuebner/libopeninv#56 (now in 51a9e58) removed the ambiguous overload, so the casts are gone and Send(id, u32buf, len) resolves cleanly.

clang-format

Brings i3LIM.cpp / leafbms.cpp / utils.cpp into line with the repo's pinned pre-commit hook (v20.1.8, --all-files), which flags pre-existing whitespace on every PR. Whitespace only, no functional change.

Testing

• ✅ Builds clean for STM32F1 (make).
• ✅ Bench-tested on a real VCU: Baud100 (100 kbps) CAN init works; Nissan Leaf inverter runs; SDO param save persists across reboot.
• ⬜ Other peripherals (chargers, BMS, other inverters) not covered by this test.

[wjcloudy]

Hardware update: flashed a build containing these changes onto a VCU — 100 kbps (Baud100) CAN init works, and a Nissan Leaf inverter runs correctly (exercises the cast Send() calls + CAN TX/RX on the new libopeninv, no regression). Still to validate before un-drafting: the web-UI SDO save / load / reset path under the new CAN-RX-ISR dispatch. Other peripherals (chargers / BMS / other inverters) not covered by this test.

[wjcloudy]

Further hardware validation: SDO param save now confirmed — changed a parameter, saved over SDO, rebooted, and the value stuck. So the CAN-RX-ISR ProcessUserSpaceSdo path (including the parm_save() flash write) works in practice. Remaining before this leaves draft is just broader peripheral coverage (chargers / BMS / other inverters).

[jsphuebner]

The ISR context for ProcessSDO wasn't fully intended, just a logical consequence. It should be fine for short-running handlers. Anything long-running should be deferred to main loop.
Having to explicitely cast to uint8_t for Send() is a bit unfortunate, API change requests welcome

[wjcloudy]

Thanks for the feedback — addressed both:

ISR context / long-running handlers: good point, and looking closer our VcuCanSdo override was actually redundant. The base CanSdo::ProcessSDO already routes user-space objects (serial 0x5000 / command 0x5002) to SdoCommands from HandleRx, so the flash save runs in the ISR via libopeninv's own default path — not our code. I've dropped the override; the PR now just removes the old GetPendingUserspaceSdo poll and uses plain CanSdo. So if the long-running SDO commands (save/load) should be deferred to the main loop, that looks like a libopeninv-level decision in ProcessSDO. Happy to add a deferring CanSdo subclass on the VCU side instead if you'd prefer that.

Send() cast: @jsphuebner took you up on the API-change offer — opened jsphuebner/libopeninv#56 (draft). Dropping the unused bool forceExt from the 2-arg Send convenience removes the ambiguity, so Send(id, u32buf, len) resolves cleanly with no cast (verified against this branch). If that lands, I'll bump the submodule here and revert all the (uint8_t) casts.

[wjcloudy]

Updated: jsphuebner/libopeninv#56 is merged, so I've moved the pin forward to 51a9e58 and dropped all 76 (uint8_t) casts — Send(id, u32buf, len) now resolves cleanly. The PR is down to the submodule bump, the CanSdo simplification, and a little pre-existing-whitespace clang-format. Builds clean without casts. Thanks @jsphuebner for the speedy merge!

[jsphuebner]

You can also drop SdoCommands::SetCanMap(cm); as that is done within CanSdo

[wjcloudy]

Good catch — done (also dropped the now-orphaned #include "sdocommands.h" since that was its only use). Pushed.

[jsphuebner]

Looks legit

[Tom-evnut]

Sorry any pushes should go to vehicle testing as per https://github.com/damienmaguire/Stm32-vcu/blob/master/CONTRIBUTING.md

[wjcloudy]

Re-targeted to Vehicle_Testing per CONTRIBUTING.md — superseded by #241 (same change, based on Vehicle_Testing).