Current implementation of ComputedShaderFlags::computeFlags(Module &M) collects a single shader flags mask for the entire module.
Change the implementation to collect shader module masks per-function of the module.
AC:
- Implement a map of function to the shader flags mask representing various properties
- Update existing tests to verify the change.
This change is not expected to include analyzing for any new shader flags other than ensuring the existing DX11_1_DoubleExtensions (viz., m_bDoubleExtensions in dxc) continues to be set correctly.