Optimize file stream metrics.

[Ted-Jiang]

**Is your feature request related to a problem or challenge?
Relate #2965
https://github.com/apache/arrow-datafusion/blob/193fc3b4549ed40a08218b744388f9211605ab53/datafusion/core/src/physical_plan/file_format/file_stream.rs#L218-L228
Before i add the cost time metric here. But there are two sub future or stream under self.poll_inner(cx)

one is a future to read parquet metadata
https://github.com/apache/arrow-datafusion/blob/193fc3b4549ed40a08218b744388f9211605ab53/datafusion/core/src/physical_plan/file_format/file_stream.rs#L162-L165

another is: for read and decode data
https://github.com/apache/arrow-datafusion/blob/193fc3b4549ed40a08218b744388f9211605ab53/datafusion/core/src/physical_plan/file_format/file_stream.rs#L177-L181

If these future are not ready will return Poll::Pending, which will make the task blocked and need someone wake itself.

So we miss the read_meta time and reading time.

Describe the solution you'd like
A clear and concise description of what you want to happen.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.