feat(sl): Add FP6/FP4 pack/unpack in C++ and Python
- Added ConvertFPtoU8 and ConvertU8toFP functions for FP6/FP4 pack/unpack support in C++.
- Implemented pack/unpack functions for 6bit and 4bit in Python.
- Added corresponding unit tests in both C++ and Python.
- Updated fp8 type names to fp8_e4m3 / fp8_e5m2 for consistency with cfloat naming.
- Moved readFromNpyFile and writeToNpyFile implementations to numpy_utils.cpp and added explicit template specializations to call the new converters.
- Reason: prevents header-level circular include/linkage issues between numpy_utils and tosa_serialization_handler.
Signed-off-by: Yuvaraj Venkatesh yuvaraj.venkatesh@arm.com
Edited by Yuvaraj Venkatesh