NAME
gh7094 - benchmark speed for keys() on empty hashes
DESCRIPTION
If you have an empty hash, the speed of keys() depends on how many keys the hash previously held.
For global hashes, getting the count for previously big hashes was substantially slower than for lexical hashes.
This test checks that the speed difference for getting the number or list of keys from an empty hash is about the same (< 25%) for lexical and global hashes, both previously big and small.
REFERENCE
This test tests against GitHub ticket #7094