Implement vulkan shaders with different data type support

Type handling preprocessor for Vulkan shaders

Start type cast functionality implementation

Update test utils and checks

Fix shader resource clearing in the vulkan device destructor

Fix optimization pragmas usages for different compilers

Replace double in shaders with float and rounding compensation algorithm

Disable vulkan device feature settings attempt

Disable atomic and float64 features for Vulkan devices

Remove using atomic float add from logsoftmax

Remove atomic add from softmax shader

Remove unused extension from shader code