remote: reduce traverse weight multiplier

[pmrowla]

In remote.cache_exists(), the traverse/full remote listing method is weighted to account for some performance hits that make it inherently slower than querying individual objects

dvc/dvc/remote/base.py

Lines 923 to 928 in 18e8f07

	# For sufficiently large remotes, traverse must be weighted to account
	# for performance overhead from large lists/sets.
	# From testing with S3, for remotes with 1M+ files, object_exists is
	# faster until len(checksums) is at least 10k~100k
	if remote_size > self.TRAVERSE_THRESHOLD_SIZE:
	traverse_weight = traverse_pages * self.TRAVERSE_WEIGHT_MULTIPLIER

The initial weight multiplier value (20) can be lowered now, due to the subsequent remote improvements that have come since the initial traverse/no traverse optimization PR