gpu-programming

After the ast generator refactor print_preprocessed is no longer used(replaced by print_preprocessed_ir). We should remove it as well as print_ast here. https://github.com/taichi-dev/taichi/blob/master/python/taichi/lang/ast/ast_transformer_utils.py#L199-L204
This task should be as simple as removing all of print_preprocessed and print_asts.
cc: @lin-hitonami

https://github.com/brendangregg/FlameGraph

The variables "lr" (learning rate), "step_size" and "step" are used in the API interchangeably. The variable "tau" is also introduced in Frechet mean. We should make this naming consistent, and for example:

• use "step_size" for integration schemes (because "step" is also used as a string to denote which type of step is taken, e.g. euler, rk4, etc)
• use "lr" for optimization schemes.