fix file stream time scanning metrics bug

[xiaoyong-z]

Describe the bug
A clear and concise description of what the bug is.
currently, file stream time scanning will be 'start()' only once, and may be 'stop()' many times. After the first calling to 'stop()', the self.start value of time scanning will be replaced to default value by 'take()'. The subsequent calling to 'stop()' make no sense.
To Reproduce
Steps to reproduce the behavior:
explain analyze select * from table;
Expected behavior
A clear and concise description of what you expected to happen.

Additional context
Add any other context about the problem here.