Improve profiles flushing using ReadRowsFrom

[cyriltovena]

This adds a flush benchmark and an optimization to reduce memory usage.

❯ benchstat before.txt after.txt                      
name      old time/op    new time/op    delta
Flush-16     400ms ± 8%     249ms ±10%  -37.76%  (p=0.008 n=5+5)

name      old alloc/op   new alloc/op   delta
Flush-16    1.03GB ± 0%    0.20GB ± 1%  -80.12%  (p=0.008 n=5+5)

name      old allocs/op  new allocs/op  delta
Flush-16     8.26k ± 1%     3.43k ± 1%  -58.54%  (p=0.008 n=5+5)

[simonswine]

We also need to call Flush() to ensure the rowGroup boundaries are still respected.

So that one segment in becomes one row group in the final file. Otherwise the ordering is incorrect. Still significant win:

❯ benchstat before.txt after2.txt
name     old time/op    new time/op    delta
Flush-8     533ms ±13%     294ms ± 6%  -44.84%  (p=0.008 n=5+5)

name     old alloc/op   new alloc/op   delta
Flush-8    1.03GB ± 0%    0.20GB ± 1%  -80.64%  (p=0.016 n=4+5)

name     old allocs/op  new allocs/op  delta
Flush-8     8.00k ± 0%     5.07k ± 1%  -36.62%  (p=0.016 n=4+5)

[simonswine]

Your bit LGTM