Ignore definition line number of functions for caching

[lhoestq]

As noticed in #1718 , when a function used for processing with map is moved inside its python file, then the change of line number causes the caching mechanism to consider it as a different function. Therefore in this case, it recomputes everything.

This is because we were not ignoring the line number definition for such functions (even though we're doing it for lambda functions).

For example this code currently prints False:

from datasets.fingerprint import Hasher

# define once
def foo(x):
    return x

h = Hasher.hash(foo)

# define a second time elsewhere
def foo(x):
    return x

print(h == Hasher.hash(foo))

I changed this by ignoring the line number for all functions.