optimize arrow batch scan

[aheev]

• improve applySemiMaskFilter perf by using lower_bound instead of linear search
• populate outputToArrowColumnIdx only once table scan

[aheev]

@adsharma could you PTAL?

[adsharma]

overall: Good improvement! I'd like to see some benchmark on what got faster

Also explain why/how setToTable() made things more efficient:

Example: one-time “bind scan projection to Arrow layout” instead of doing the mapping repeatedly.

[aheev]

overall: Good improvement! I'd like to see some benchmark on what got faster

Also explain why/how setToTable() made things more efficient:

Example: one-time “bind scan projection to Arrow layout” instead of doing the mapping repeatedly.

There's shouldn't be significant improvement. It's a short(#columns are usually small) for-loop. I just happened to notice redundant re-population of outputToArrowColumnIdx every morsel scan