Make safe_filename() more efficient

The current implementation is using re.sub() to replace a set of characters in filenames.

After a small benchmark, and given the huge number of filenames Drive is is dealing with, this micro optimization may be interesting.

from timeit import repeat

setup1 = r"""
import re
pattern = re.compile(r'([aetuo])')
value = "azertyuiop"
"""

setup2 = """
repmap = {ord(c): "-" for c in "aetuo"}
value = "azertyuiop"
"""

print(repeat("re.sub(pattern, '-', value)", setup1, number=100000)[1])
print(repeat("value.translate(repmap)", setup2, number=100000)[1])

The script first test the current implementation using re.sub().
The second is using a potential better implementation, using str.translate().

Results:

re.sub()       : 0.23668296402320266
str.translate(): 0.05809920001775026

This is a significative improvement on the performance level.