boruta_py | Python implementations of the Boruta all-relevant feature | Machine Learning library

New to Machine learning so please bear with me, thanks!
I have three questions to ask, so it would helpful if you mention the question number while answering.

so I want to perform feature selection for my training data before i apply the machine learning algorithm. I will use the same data set to run on many different ML algorithms to decide what is best so it will be more efficient if i can just do feature selection once and pass the new data set to the various algorithms.
Note : I am coding in Python3 and I'm going to use BorutaPy for my feature selection. [https://github.com/scikit-learn-contrib/boruta_py]

Question 1)
do i need to know what algorithm i'm using before performing feature selection? or can i just perform my feature selection and then use whatever algorithm ,ie; is feature selection dependent on the type of algorithm used?

Question 2)
can i perform the same feature selection for regression and classification problems?

Question 3)
Instead of everything mentioned above, is it best to use regularization for the regression problems and perform feature selection for the classification problems?

Thank you!