Fixing parquet coalescing of reads

[hyperbolic2346]

When digging through the new chunk writer for parquet pipelined reads, I noticed that the code changed and the read coalescing was no longer happening. Through some checking, I found that tests were issuing multiple smaller reads, specifically a read per rowgroup instead of a single read for multiple rowgroups. With this change, there is a single read per file.

I believe this is an unintentional change and a departure from the previous behavior.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[hyperbolic2346]

/merge