feat: read all bytes out before do async bench

[jiacai2050]

Which issue does this PR close?

Closes #

Rationale for this change

In parquet sync read, file is read out before iterate record batch, this PR update async read bench, to let it have same logic with sync version

What changes are included in this PR?

Are there any user-facing changes?

How does this change test

Not required

A benchmark for parquet file with 103951966 rows

[2022-11-14T09:03:56Z INFO  benchmarks::parquet_bench]                                                                                                               
ParquetBench Async total rows of sst:103951966, total batch num:12690,
                open cost:790.76716ms, filter cost:149.713098ms, iter cost:9.603906449s

[2022-11-14T09:31:39Z INFO  benchmarks::parquet_bench] 
ParquetBench Sync total rows of sst:103951966, total batch num:12690,
                open cost:969.653961ms, filter cost:165.7483ms, iter cost:9.981503914s

[ShiKaiWi]

LGTM