[rocBLAS][Tensile] use kernel ISA or print error, CurrentISA should be benchmark use only

[TorreZuk]

• revise ISA detection as bootstrap only for benchmarking
• fix kernel scheduling and generation to error if no ISA (but continue)
• consider runtime error if no kernel ISA ?

[Copilot]

Pull Request Overview

This PR revises ISA (Instruction Set Architecture) detection to use kernel-specific ISA or print errors when unavailable, changing CurrentISA to be primarily for benchmarking bootstrap purposes.

• Modifies kernel initialization to prefer kernel-specific ISA over global CurrentISA
• Updates ISA detection logic to be more robust with fallback mechanisms
• Improves error messaging and warning output for ISA-related issues

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
KernelWriterAssembly.py	Updates kernel initialization to check for kernel ISA first, with error messaging for fallback
KernelWriter.py	Adds ISA checking logic to kernel scheduling with error output for missing kernel ISA
Common.py	Renames detection function, improves ISA detection flow, and updates warning messages

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[TorreZuk]

./library/src/build_tmp/TENSILE/assembly/asm-new.sh also has issues that must be resolved as it generated for the bootstrap ISA which would not be applicable. No change yet but can add to this PR

[TorreZuk]

Latest work is to replace the CurrentISA global use for kernel generation so will close this PR. It highlighted the incorrect usage of the global but don't expect we are going this route.