Rework of -target/-mcpu/-mfpu options for LLVM CLANG

[chentang16]

Due to various issues reported so far:

• [Bug]: The -mfpu option (CLANG_MFPU) is missing when CLANG toolchain is selected cmsis-toolbox#583
• fatal error: error in backend: arm/arm-toolchain#861

the -target/-mcpu/-mfpu options in CLANG.17.0.1.cmake have been reworked based on the ATfE manual pages:

• https://developer.arm.com/documentation/107976/22-1-0/Compile-for-a-specific-Arm-architecture-or-Arm-processor/-mcpu-command-line-options-for-M-profile-processors?lang=en
• https://developer.arm.com/documentation/107976/22-1-0/Compile-for-a-specific-Arm-architecture-or-Arm-processor/-mcpu-command-line-options-for-A-profile-processors
• https://developer.arm.com/documentation/107976/22-1-0/Compile-for-a-specific-Arm-architecture-or-Arm-processor/-mcpu-command-line-options-for-R-profile-processors

Fixes

• [mitigation] CLANG commandline option mcpu=cortex-m55 -mfpu=fpv5-sp-d16 can lead to fatal error in backend #2491
• [Bug]: -mcpu/-mfpu options in CLANG.17.0.1.cmake need to be reworked cmsis-toolbox#618

Changes

• CLANG_MCPU
• CLANG_MFPU
• CLANG_TARGET
• CLANG_ARCH
  options for Cortex-M/A/R covered by CMSIS-Toolbox.

Changes have been tested with all Cortex-M built targets in https://github.com/Arm-Examples/AVH-Hello
No more crash issue tested with https://github.com/Arm-Examples/cmsis-mlek-examples
No more hardfault issue tested with Tensorflow_LiteRT_HelloWorld on STM32F7-Disco

Checklist

• 🤖 This change is covered by unit tests (if applicable).
• 🤹 Manual testing has been performed (if necessary).
• 🛡️ Security impacts have been considered (if relevant).
• 📖 Documentation updates are complete (if required).
• 🧠 Third-party dependencies and TPIP updated (if required).

[Copilot]

Pull request overview

This pull request reworks how the LLVM Embedded Toolchain (Clang) CPU/arch-related flags are derived in CLANG.17.0.1.cmake, aiming to resolve reported backend crashes/hardfaults by adjusting the use of -target, -mcpu, and -mfpu according to Arm toolchain guidance.

Changes:

• Reworked per-CPU mappings to prefer encoding FP/MVE/DSP capabilities into -mcpu feature strings and, in many cases, omit -mfpu.
• Adjusted target triple selection logic (notably for Cortex-A/M/R vs other arches) and added conditional handling for -mabi.
• Added validation/error branches for unsupported FPU/MVE combinations for Cortex-M52/M55/M85 and related CPUs.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 65.24%. Comparing base (859cb6c) to head (84d3dce).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2498   +/-   ##
=======================================
  Coverage   65.24%   65.24%           
=======================================
  Files         147      147           
  Lines       26675    26675           
  Branches    16166    16166           
=======================================
  Hits        17405    17405           
  Misses       7074     7074           
  Partials     2196     2196           

Flag	Coverage Δ
buildmgr-cov	79.87% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[chentang16]

The incorrect FP_FVE condition check in the first commit was due to
Open-CMSIS-Pack/cbuild2cmake#485

[brondani]

I see a couple of shortcomings that have a wider scope and could/should be fixed as part of the related issues:

1. INT_MVE should be MVE, related to:
   Open-CMSIS-Pack/cbuild2cmake#485
2. PACBTI for Cortex-M52, M55, M85, related to:
   #2285

The rest looks good to me.